Challenges and review of goal-oriented requirements engineering based competitive non-functional requirements analysis

Abstract

Modelling and analysis in software system development can be especially challenging in early requirements engineering (RE), where high-level system non-functional requirements are discovered. In the early stage, hard to measure non-functional requirements are critical; understanding the interactions between systems and stakeholders is key to system success. Goal-oriented requirements engineering (GORE) has been successful in dealing with the issues that may arise during the analysis of requirements. While assisting in the analysis of requirements, $i^{*}$ goal model is the only framework available among the many GORE models, emphasising socio-technical domains such as stakeholders/actors/players, goals/objectives, dependencies and design options/alternatives. Most current approaches to goal-model analysis use quantitative methods or formal information that is hard to gather in early RE, or produce analysis results automatically over models. In real-time competitive applications, the goals of various stakeholders are conflicting in complex systems. Also, each of the system goals have various alternative design options for the systems and optimal selection of goal-oriented requirements faces several challenges in requirements-based engineering. Hence, effective decision-making frameworks are necessary to capture the real issues to achieve multi-objective optimisation of interdependent actors. To obtain an optimum strategy for interdependent actors in the $i^{*}$ goal model must balance the opposing goals reciprocally. To achieve this, the model needs to go beyond the analytical decision-making tools such as sensitivity analysis tasks, cost-effective analysis process, game-theoretic concepts and analytical hierarchical process. To address these requirements, this paper discusses the design of novel frameworks for an agent-based goal model analysis in requirements engineering. The objective of this paper is to provide a brief and comprehensive review of the major efforts undertaken along this line of research. In this paper we have prepared literature review of the concepts, terminology, significance and techniques of Goal oriented requirements engineering in the context of non-functional requirements analysis.

Keywords

Requirements engineering goal models multi-objective optimisation decision-making non-functional requirements

ï»¿

1. Introduction

In a global perspective, due to the rapid growth of technology, software systems tend to make greater developments in the open, distributed, decentralised and integrated environments. This influences fields such as e-commerce, e-learning, e-banking, e-healthcare, e-forecasting and so on, because of the advancement from anywhere, at any time and in any form. A newer technology called agent-oriented technology is adopted due to the demand for enabling the system to adapt to environmental variations and higher user requirements. Agent-oriented software engineering has received more research attention [74, 38, 68, 27] in the field of software engineering. The following software phases should satisfy the entire software development cycle for the development of any software system, such as design, requirements, evaluation and verification, evolution, deployment, maintenance and so on. Among these factors, requirements engineering (RE) is a very challenging task in agent-based software technology.

Requirements engineering utilises systematic techniques to satisfy the system requirements in terms of completeness, consistency and relevance factors. This type of engineering is utilised by stakeholders, users, developers and analysts to determine each other user’s interests, and examines other options for decision making about the systems to be implemented. Based on the software requirements, the system is categorised into functional requirements (FR) or non-functional requirements (NFR). Behavioural/functional requirements define the system’s components or functions; whereas non-functional requirements deal with system operation features rather than particular behaviour. Some of the features include usability, while integrity and security of softgoals have more effects on software systems rather than the goals.

Several approaches such as structured modelling, object-oriented, use case, conceptual entity-relationship modelling and goal-oriented approaches have been proposed in the RE literature. Compared with traditional models, goal-oriented requirements engineering (GORE) is appropriate for requirements analysis in terms of the software development cycle of NFR and the computation of alternatives [49]. The GORE approach examines, reveals and expresses stakeholders’ goals, which leads to system and software requirements. Several other approaches are employed for the requirements analysis in the software cycle and traceability analysis for the later stage.

Goal models are utilised by analysts for the estimation of goals, evaluation of design alternatives, system design selection, risk analysis and prioritisation of the system requirements. For the purpose of implementation, the best design is selected from different design options. Hence, in the evaluation of design alternatives, different design alternatives are determined, and the best is selected based on selection criteria. In the previous quantitative and qualitative models, softgoals are used as the evaluation criteria [48]. During the validation stage, the values are propagated either by forward (that is, from bottom to top softgoals), or backward (that is, from top to bottom softgoals). Hence, the satisfaction levels of the softgoals and the satisfaction level concerning the selection of top softgoals are estimated based on the selected design alternatives.

Qualitative labels such as partially satisfied, satisfied, partially denied, denied, conflict and unknown are used in the qualitative models. This type of approach has the major drawback of ambiguity problems, which arise due to the same label, and the goal receives an unknown or conflict label during the decision-making process. Therefore, these issues are analysed to improve the conflicting requirements with software development [44]. In qualitative analysis, the numbers are used instead of qualitative labels. Thus, definite numbers are assigned as per the stakeholders’ requirements because the analysis may consider dissimilar stakeholders with different requirements as per the requirements’ analysis task. This is because the discrete stakeholders have various knowledge and training levels [70]. Also, linguistic factors such as minimum cost and maximum profit are considered by the stakeholders to establish communication based on their requirements preferences. However, the representation of these terms in definite numbers is also challenging. Due to those complications, it is necessary to design a novel approach for the goal analysis. When comparing with NFR, knowledge acquisition in automated space (KAOS), Tropos, and $i^{*}$ , goal-oriented requirements language (GRL) goal models show various actors’ dependencies on other actors for achieving the goals. In the decision-making process, these inter-dependencies are very influential for other design alternatives [5].

In the literature [46], the actor dependencies and formalisation approaches have been discussed according to the RE process. Due to incomplete information, the requirements needed for the evaluation by the input data are imprecise. To address RE problems, the analysis deals with more than one goal, which is subjected to multiple-goal models in the real-time environment. However, the priorities of these multiple goals may differ and might be conflicting, but have to be considered simultaneously. Also, scalability is an issue for designing the goals, and when the goal size increases, it is difficult to assign values and complications occur during the decision-making process [75, 30, 71]. Therefore, it is necessary for automation in the goal analysis procedure.

In real-time competitive applications, the goals of various stakeholders are conflicting in complex systems. Also, each of the system goals have various alternative design options for the systems. In addition, optimal selection of goal selection faces several challenges in requirements-based engineering. During the decision-making process, several concerns are considered for the estimation of interdependent interactions among actors. Thus, an effective framework is necessary for learning the issues to attain multi-objective optimisation [51]. The decision-making process in the realistic approach has made the process go beyond the analytical tools such as sensitivity analysis tasks, cost-effective analysis process, game-theoretic concepts and analytical hierarchical process.

In respect of these requirements, this article addresses the design issues of a novel framework for the agent-based goal model analysis in requirements engineering. The major objectives for the proposed work are summarised herewith. The paper aims to address design issues in decision-making approach, algorithms and tools to facilitate the reasoning and analysis for the early agent-goal requirements. Further, for the interdependent actors, the selection of optimal alternatives by balancing the multiple conflicting/opposing objectives reciprocally, providing analytical model analysis and reasoning and performing multi-objective functions are also discussed in this paper. The main purpose of this paper is to offer a concise and thorough overview of the key research initiatives in this area. The core ideas and terminology of RE, GORE, and decision-making analytical methodologies are presented in Section 2. Section 3 discusses some of the unique issues of adopting agent-based goal models for early requirements analysis. Section 4 presents existing methodological approaches to goal model analysis and non-functional requirement reasoning. Section 5 specifies future scope and the paper concludes with a brief summary.

2. Research issues related to competitive non-functional requirements analysis

Requirements engineering (RE) is frequently regarded as a front-end activity in the software and systems development process. RE is, as its name proposes, the engineering discipline of establishing user requirements and specifying software systems. During the development stage, these requirements are utilised by the software developer in defining what functionality needs to be developed and delivered [26, 4, 9]. Various requirements include the specification of features for the system, based on system behaviour, which represents the general property, as well as constraints, and algorithmic data for system engineers. Generally, the requirements are categorised as functional and non-functional/domain requirements. Functional requirements refer to the system function and its components, whereas non-functional requirements check the system operations rather than the system behaviour. Some of the non-functional requirements factors (such as integrity, usability and security) have a greater effect than the functional requirements of the overall software system.

In software engineering, goals are utilised to estimate the early non-functional system requirements [24, 65, 57, 40]. Numerous approaches are proposed earlier in RE, such as goal-oriented, conceptual entity-relationship modelling, use-case approaches and structured modelling. GORE and NFR models are suitable for examining and revealing the stakeholders’ goals, which lead to the system and software requirements. At the final phase of the software cycle, other approaches are more suitable for requirements analysis and traceability evaluation between requirements and implementation tasks. GORE deals with the goals for documenting, negotiating, structuring and requirements reconstruction [66, 13]. Here, stakeholders’ objectives define the goals presented by the system where the software is to be integrated under its environment. Goals are specified based on AND-OR representation that defines the functional as well as NFR, and is classified from higher to lower values. Some of the properties of system goals include fault tolerance, survivability, and specific safety achieved for high assurance organisations. The factors affecting such high assurance models are goal modelling and reasoning tasks. Various frameworks of GORE in literature are Tropos [11], KAOS [15], NFR [14], $i^{*}$ framework [75], GRL [5] and attributed goal-oriented requirements analysis (AGORA) [33]. Readers are directed to these references for detail.

To support qualitative reasoning of goals, goal models are designed during requirements engineering. This model consists of an AND-OR graph that illustrates how the high-level goals are contributed by the lower ones, and vice versa [67]. Also, the satisfaction level is estimated by the analyst based on goal models, which also evaluates design alternatives, selects system design, analysis of risk and requirements prioritisation. During the evaluation stage of selection criteria, alternative designs are evaluated, and explore various other designs to select the best one. In the goal model, softgoals are utilised as the evaluation criteria in the earlier qualitative and quantitative estimation approaches [49, 48]. In the RE literature, several qualitative and quantitative models are proposed to support the goal analysis process [72, 41, 45, 32, 19, 5, 6]. In the goal model evaluation stage, the qualitative/quantitative measures are transmitted from the bottom to the top softgoal. Also, the depiction of the contribution of goals to softgoals is carried out by assigning the weight values in terms of both positive and negative values. In the goal model, some of the qualitative labels are used for the allocation to the nodes. To determine the satisfaction level, these labels are propagated through link paths for the achievement of the goal. The decision-making process becomes difficult when two or more different options with the same label are considered. In some cases, when an option has an “undetermined” or U label, the decision process is uncertain. To solve these problems of qualitative reasoning, quantitative reasoning [61] has been proposed as a measurable specification in the RE literature. The evaluations are performed to define the contribution from softgoals to goals. In the quantitative model, the labels are transmitted via links to determine the level to which the goal has been satisfied. Based on using qualitative and quantitative labels, several research models are considered relating to goal achievement [1, 5, 19, 32, 56].

Let’s focus attention to available analytical decision-making techniques from the literature and look at how they are effectively used in RE. Game theory is a powerful interdisciplinary tool that helps researchers to analyse complex situations in multi-agent systems and RE design challenges [73, 62, 64]. It has been mainly applied in the fields of mathematics and economics that effectively describes the relationships between decision-makers. The main idea behind this theory is the generation of an ideal solution under specific conditions under the assumption that the players are rational and that they perform tasks based on their interests [36]. In RE design, a non-cooperative game theory is introduced by Yazdania et al. [73] which enhances the sub-optimal performance of the project design. During the development stage, the complex system and its requirements are classified into subsystems and the subsystem-level requirements which is then fulfilled by each design team. The performance of the entire system depends on the allocation of the resources shared by the subsystems. The requirements specification should be satisfied by each team, which leads to the reduced decomposition at system-level requirements and also affects the design alternatives. Thus, an optimal design [58, 60, 28] is not achieved as a result. Hence, to derive an optimal result output, a non-cooperative game theory is applied for the theoretical analysis of RE design. However, this technique cannot be applied to goal analysis while it employs game theory in requirements design-based engineering.

For the analysis of strategic decisions, objectively, the CEA tool has been developed which investigates the costs, productivity, performance and system efficiency. In this approach, the cost investigation is estimated in terms of money and significance using non-monetary terms. For instance, the natural units are computed in physical terms, including lives saved, cases cured, complications and so on [10, 50, 63]. This process results in an accurate cost prediction and the expected result [53]. Hence, a trade-off analysis is involved while choosing between the available options. For the performances of each alternative, gradual expenditures, the total costs and the effects are also computed. The result of CEA is described by a cost-effectiveness ratio (CER); the denominator describes the advantage of selecting particular alternatives; while the numerator defines the cost related to this benefit. This tool helps decision-makers with the selection of the best design alternative that satisfies technical and financial requirements. After the calculation of CERs based on the societal perspective, an alternative design is selected within constraints executed by available resources and leads to the lowest cost per effectiveness. The CEA offers a simple as well as a critical contribution by the computation of overall costs and outcomes. This helps in the selection of design alternatives for a given issue. If the overall cost and the result is misrepresented, then several alternatives cause complications, depending on its application. Thus, the need for the data to be accurate becomes critical because the decision making is directly proportional to it. The inaccurate choice of data leads to the selection of wrong alternatives and causes an impaired decision-making process. Thus, an accurate selection of information helps to derive the best alternatives, which in turn leads to an effective CER. Based on this process, the best cost-effective alternatives with low costs are predicted. In this research work, a CEA-based multi-objective optimisation and the analysis of NFR in the $i^{*}$ goal model are introduced.

A multi-objective decision-making model was developed by Thomas T Saaty in 1972 based on pairwise assessments between the design alternatives [54]. In this article, a hybrid quantitative AHP based satisfaction propagation model is also proposed. For the pairwise comparison process, this framework is introduced after the goals’ decomposition (goals into sub-goals between the non-functional requirements). For the prioritisation and the preference requirements evaluation, Liaskos et al. [41] present a model that obtains the important requirements while satisfying the best preference priorities and requirements. This model distinguishes the mandatory goals from preference goals and suggests a model for finding other ways to accomplish mandatory and the fulfilment/non-fulfilment of preference goals. However, this technique cannot be applied to real time situations where goals are of opposing nature. For requirements prioritisation, Sadiq and Jain [55] propose a prioritisation of requirements model based on the analysis of AHP. For group decision-making, this model employs a fuzzy model and the set of prioritised requirements obtained by a binary sort tree method. Also, the weight values based on the AHP pair-wise comparison model are assigned to locate the list of prioritised requirements. The integration of experts’ preferences with group preferences is performed by fuzzy preference relation. The disadvantage of this approach is that, for the demonstration of this approach, a small number of requirements and criteria are used, and high level mathematical knowledge is required for the prioritisation of goals.

3. Challenges in agent-based goal model reasoning for early requirements engineering

Some of the specific challenges for early requirements analysis have been identified, such as non-functional importance, struggling for adequate accuracy and completeness. These challenges play an important role in making effective reasoning and analysis procedures using agent-based goal models in

Table 1
Analysis methods in GORE

Qualitative Methods
References	Approaches	Strengths	Limitations
[67], [25], [32], [48]	Applied qualitative labels like satisfied, partially satisfied, denied, unkown, conflict and ++, +, – or - in the propagation algorithms	Easy and simple goal analysis	Vague reasoning, Difficult to implement in large systems, requires strong mathematical reasoning, decision making leads to ambiguity
[16]	Applied qualitative labels like categorical data and pictures, based on human behaviour	Helps to quantify human characteristics, investigate the complexity of the problem	Results are softer and fuzzier than quantitative results, hard to summarize and simplify
Quantitative Methods
References	Approaches	Strengths	Limitations
[44], [5], [41], [17]	Applied quantitative labels like numeric values, probabilistic values in the propagation algorithms, single objective and multi-objective optimisation goal analysis	Avoids ambiguity in goal analysis	Requires certain structures to be satisfied for goal analysis, Difficult to implement in large systems, requires strong mathematical reasoning
[39]	Applied quantitative labels like data obtained from samples, based on dependent and independent variables	Helps to reduce the complexity of the problem, Results are of great accuracy, avoids personal bias	Results are quantitative results, no real world application, no human perception

Table 2

Comparative analysis of goal reasoning methods in GORE – NFR model

References	Approach	Analysis type	Optimisation	Tool implementation	Sensitivity analysis	Game theory	AHP	CEA	Probabilistic approach
NFR model
[4]	Non-functional requirements framework: A mathematical programming approach	Qualitative	Yes	No	Yes	No	No	No	No
[28]	Optimal Requirements Dependent Model-Driven Agent Development	Quantitative	No	No	No	No	No	No	No
[44]	Utilizing TOPSIS: A multi criteria decision analysis technique for non-functional requirements conflicts	Quantitative	No	No	No	No	No	No	No

Table 3

Comparative analysis of goal reasoning methods in GORE – KAOS and GRL models

References	Approach	Analysis type	Optimisation	Tool implementation	Sensitivity analysis	Game theory	AHP	CEA	Probabilistic approach
$i^{*}$ model
[32]	Evaluating goal achievement in enterprise modeling – an interactive procedure and experiences	Qualitative	No	Yes	No	No	No	No	No
[41]	Integrating preferences into goal models for requirements engineering	Quantitative	No	No	Yes	No	yes	No	No
[42]	Representing and reasoning about preferences in RE	Quantitative	No	No	Yes	No	Yes	No	No
[17]	On the quantitative analysis of agent-oriented models	Quantitative	No	No	No	No	No	No	No

Table 4

Comparative analysis of goal reasoning methods in GORE – $i^{*}$ model

the analysis of early requirements [59]. In this section, several challenges in agent-based goal model reasoning for early RE are outlined below.

3.1 Complexity in model analysis

Agent-based goal models can become too complex to be manually reasoned and analysed. One way to ease the complexity of analysis over goal models is to automate the analysis process. However, some level of automation is required to support model structure and content analysis. Although the appropriate level of support is found several times, the level of automation is difficult. Too much automation fails to account for the inherent in-expressiveness and incompleteness model in early RE and may limit the role of the modeller or stakeholders. Also, insufficient automation can lead to a process that is too tedious or time consuming to realistically complete and can produce inconsistent and difficult to reinterpret results later. It can also be challenging to understand the results of the analysis over a complex model. Methods that aid in understanding the model or analysis would help such analysis be more accessible to stakeholders with limited time.

3.2 Completeness of modelling

According to the high-level social nature of early RE models, it can be argued that models are never complete in the same way as models used for other purposes. In other modelling contexts, there are often clear requirements for model inclusion. For example, the construction of entity-relationship diagrams (ERD) entities involved in a focus system is very limited in the existing system. When modelling the early requirements, the models represent only a complicated and system interconnected part of the network with people. There are lots of stakeholders, more interacting systems, more goals, dependencies and contributions that can be added to the models. The difficulty of exploiting the modelling lies in two facts:

(a)
knowing where to stop modelling
(b)
knowing if the amount of information collected is sufficient to support useful analysis, reasoning and understanding of the domain.

3.3 Accuracy of model analysis

Model accuracy issues are similar to model completeness issues. To generate a model, the interactions among the social needs are difficult to characterise accurately for analysis and understanding. The model can be argued from a constructionist viewpoint when the model is correct or not. However, the models have sufficient collection modellers from the collective viewpoint regarding the domain. Hence, improving accuracy and generating sufficient accuracy are the challenges in early RE [34].

3.4 Understanding of the domain

One of the primary aims of modelling, reasoning and analysis is to increase the understanding of the domain in early RE. In the early requirements process, the large, complex and socio-technical organisation are the challenges to understand the domain. Practically, stakeholders want to focus on the technical details without receiving the motivations or conflicting goals that can frequently prevent the technical choices underlying those successful requirements. For an effective technical intervention, the sufficient level is specified by early RE models and analysis. Therefore, it is difficult to know the depth of all technical details for certain project time constraints. However, understanding the domain is complex to balance the requirements in the domain.

3.5 Flexibility in modelling

Goal models have been developed specifically to addresses the level of flexibility and in-expressiveness to facilitate the explicit consideration of high-level NFR and social requirements (for example, customer satisfaction, company branding). These requirements are difficult to quantify and formalise, but should be recognised and considered in the early stage analysis. Although clear measures can be assigned to requirements, it is difficult to determine how to combine measures at different scales in an integrated and accurate way. The critical system involves early, inexpressive and ambiguous representations. Once key decisions are obtained, the scope of the critical system becomes restricted or partially sufficient, and its resources can be measured or formalised for later review. However, the goal models of early RE will support reasoning about critical system requirements.

3.6 Involvement of stakeholders

An active part of the early RE process is important for stakeholders. Stakeholders already provide the collected information to validate their needs and interactions. The inherent incomplete and inaccurate nature of goal models is especially important to encourage continuous iteration as a means of stakeholder validation. Depending on the domain, system users may be reluctant to support upcoming changes for technical or political reasons. The goal modelling and analysis approach should have a relatively low learning curve. They should have a reasonably transparent rationale and functionality for users with a given shortage of time for stakeholders. Furthermore, stakeholder participation can induce “buy-in” or a sense of ownership in project goals and planned changes to improve the accuracy and completeness of the models in the early system analysis. Stakeholder scheduling makes it difficult to access their time when it is busy. In practice, the stakeholders often have difficulty understanding models due to the complexity of analysis. To solve this issue, an effective method is needed to involve the stakeholder as much as possible for the modelling and analysis process.

3.7 Analysis and reasoning on modelling

Without analysis, the process of generating an agent-based goal model can be useful for understanding and agreement. Such models should support reasoning and analysis as much as possible to increase profit for the time invested. In other words, modellers should be able to use the models to answer different types of useful domain questions. Although the structure of goal models allows multiple procedures, most procedures require the addition of specific formal or quantitative information to the goal models [40]. This information is needed to encode the model. To facilitate different types of reasoning and analysis, there is a trade-off to analyse more specific information, including the time and difficulty in finding this type of information to generate the goal model.

3.8 Usability and selection of decision-making methodologies for goal analysis

In early RE, the system decisions are prepared by group consensus that implicitly includes and possibly guides without documentation. Various analysis procedures collect some information to make key initial decisions from the usability of goal models. However, the model and analysis may not achieve clear results because certain decisions were made based on judgments expressed about contentious areas and assumptions support the stakeholder choices in the model. Also, it is difficult to know what information needs to be retrieved and associated with the goal-model-aided decision process. In this, the modelling process becomes slightly complex due to a large amount of information and stakeholder leaves their own or other choices due to small amount of information.

For agent-goal model analysis and decision-making process, several existing methods have been introduced, but there is little work focused on the practical usability of existing methodologies. Such methodologies are usable, and it is not clear to express the requirements. The complex analysis methodology [24] is employed by expanding goal model syntax [22] in several goal models approaches. These methodologies mentioned the objective of simplicity and usability requirements. For goal model analysis, a wide range of questions are analysed from the available methodologies below.

1.
What procedure would the stakeholder and modeller choose if the performance analysis is based on an early RE content?
2.
How can users choose the methodology for analysing the possible goal model suited to their needs?

Therefore, understanding benefits, capabilities, and costs are the challenging issues to make the goal model analysis techniques possible in RE.
4. Approaches to goal model analysis and reasoning of non-functional requirements

4.1 Conceptual foundation

In the model analysis, several approaches are identified based on the goal of denial or satisfaction. At the initial stage of the procedure, the model initiates assigned value to propagate either forward or backward for an alternative reflection/question using model links [14, 24, 40, 5]. These procedures analyse several alternatives of forward and backward questions, such as, “Is it possible to satisfy certain goals?” or “What is the alternative impact of forward? If possible, which alternative would the model meet these goals?” (that is, forward or backward). Typically, some satisfaction procedures represent analysis results using partially satisfied, satisfied, denied and partially denied in terms of qualitative labels [5, 24, 14]. To deal with quantitative analysis, the probability of the goal can be represented in the degree of satisfaction/denial [5] or satisfied/denied [40, 24] for several procedures. On the other hand, the remaining procedures have only one or two values to produce binary results, whether it is typically satisfied or not.

Among these approaches, the main distinguishing features of multiple incoming values are resolved for the goal model. To various degrees, the contribution links represent positive and negative consequences from the goal models. Different types of contributions receive various goal strength in the formation of positive and/or negative simultaneously. Such situations could be resolved by separating negative and positive evidence regardless of whether it is unnecessary or not for some approaches [24]. Other procedures employ predefined qualitative or quantitative rules to combine multiple values [5, 40]. Besides, the interactive procedures are used to resolve partial or conflicting evidence with the use of human intervention based on domain knowledge [14]. These procedures utilise construct metrics to measure quality in the domain of security, vulnerability and efficiency [19].

The procedures analyse several questions such as, “How does the specific alternative for a specific stakeholder become risky?” or “How does the system become secure by representing the model?” To illustrate, Franch et al. [18] present classifications and weights of actors over $i^{*}$ strategic dependency (SD) models to measure global or local metrics. This approach [17] is expanded to develop the framework that allows interactive metric calculation and give an effort over $i^{*}$ strategic rationale models.

The procedure analyses several questions, such as, “What is the best action plans on certain requirements?” or “What design alternatives need to be taken to meet the goals?” In this instance, the Bryl et al. [12] analyse plans to determine satisfactory delegations based on dependencies through the goal model in the social network. To fully exploit the potential of the goal model, the evaluation is measured in terms of cost metrics and use similar metrics [18]. Many approaches use construct metrics to represent the goal model and allow temporal information for implementation over the network [69, 22]. In this case, the simulated results are tested to obtain unexpected or interesting properties in a particular scenario. These procedures analyse different questions such as, “What particular alternative is selected if the analysis is obtained?”. Further methods provide additional information to perform enhanced ways over the models and ask different questions to the users, such as, “What makes the model consistent?” or “What model is particularly possible to achieve a goal?”. To illustrate, the desired constraints are represented to add the model for linear-time temporal logic statements and transform $i^{*}$ models to formal TROPOS using first-order logic. Then the model validates the property and verifies the illustration for consistency, Fuxman et al. [20, 21]. Instead of analysis, several methods focused on model construction methods to generate models with the help of modellers [8, 29, 40]. These methods involve process re-engineering $i^{*}$ Method (PRiM) in which $i^{*}$ models employ complementary artifacts (context models and Data Flow Diagrams) and separate processes from intentional content that is systematically constructed for model analysis and reasoning [29].

4.2 Identification of research gaps in goal model reasoning

For early RE, several challenges that appeared in agent-based goal analysis are outlined in Section 1.2. Existing methods do not explicitly define early RE challenges. Indeed, most of the works are focused primarily on the analytical power offered by their procedures. Several procedures address the issues faced externally and motivate the scope of goal model analysis for early RE. For example, the system operation could be analysed in detail for simulation. The contributions of existing work challenges are discussed below.

4.2.1 Complexity in model analysis

Some level of automation is needed to support model structure and content analysis over complex goal models. In existing procedures, the fully automated process takes particular content as input, and the majority of the model content is used to automatically produce results for goal model analysis. However, the fully automated analysis makes it difficult for user-analysts and stakeholders to trust the results due to incompleteness and inaccuracy. Also, the analysis procedure can be difficult to understand (how the results are achieved), and validation of the results is challenging due to the transparency and complexity of the model. For this purpose, that fully automated analysis is observed as being not ideal for early stage RE analysis. Some analysis procedures have introduced interactive components that allow the modellers to get involved at various points [7, 18]. For example, analyst intervention is used to promote or demote partial evidence or to decide whether the evidence conflicts with the NFR procedure. Although it is useful for allowing the modeller to intervene in the modelling process with their domain knowledge, the restrictions on user intervention of all values must be specifically promoted. This leads to a limited loss of information. Hence, existing approaches have not focused on supporting knowledge of analysis results over complex models.

4.2.2 Completeness of modelling

For the best knowledge of the domain, the goal model analysis procedures do not address model completeness issues. Most of the procedures are processed with model assumptions that can be analysed accurately and completely. Existing procedures focus on analytical power and cannot determine gaps in the model acquisition of the knowledge. Although the analysis procedure can potentially reveal important missing information when the results are examined by stakeholders, finding errors is particularly difficult when analysing the “future” situation. Then it can be useful for modellers to re-examine and question the completeness of the model, especially fragments that have a high degree of uncertainty when analysing early RE models.

Automatic analysis procedures may not motivate a model review. Finding and examining these areas in an ad hoc way may be challenging for modellers, especially if they have already spent a lot of time generating the model and have reached an agreement based on its contents. As a part of the model analysis, the required methods help to deal with important model fragments, finding potential errors and improving the quality of the model. The suggested steps achieve completeness with the help of modellers, and also verify other types of models for model construction using different approaches [29]. This can be particularly helpful while constructing the model initially. The generated steps must become a complete model based on the complex nature of agent-based goal models. To improve the completeness of the model, the agent-based goal model focuses on intention and controls against non-intentional models. Hence, further approaches are needed to increase confidence in the completeness of goal-oriented models and help to achieve model stability.

4.2.3 Accuracy of model analysis

For any analysis procedure, modellers examine and reveal inaccurate parts corresponding to the completeness of the analysis of a model. But existing methods focus on analysing the domain represented by the model and validate that either the model is correct or not. Initially, the model is generated using the basic “sanity” tests, especially when the correctness of the model has to be verified using analysis procedures. To this end, the user guide requires the use of goal analysis procedures, which includes several analysis questions to ask about the newly finished models. Similar to model completeness, the modellers continually re-examine the model key areas to determine the model inaccuracies in this case. This process will not be supported for fully automated analysis. Hence, there is a need to determine a more appropriate procedure to balance between automation and intervention for goal models used in early RE.

4.2.4 Understanding of the domain

To understand the domain, any procedure can help provide analysis results over the domain model. However, existing procedures do not explicitly aim to increase or improve domain understanding, but instead focus on answering specific questions, often selecting the best design alternatives, without ensuring the selection criteria is sufficiently accurate or complete.

4.2.5 Flexibility in modelling

Several procedures support reasoning based on flexible and inexpressive models through the use of simple and qualitative labels [5, 14, 74]. Such labels can be applied and propagated using a mix of automated rules and stakeholder judgment without forcing users to formalise or quantify high-level model concepts. However, other procedures use a quantitative interpretation over these informal concepts, assuming that the numbers are meaningful; that is, customer satisfaction of 0.7 means that this goal is satisfied on a scale of 7/10 [33, 5, 24]. On the other hand, other procedures require that the model have a precise formal or quantitative label before undergoing analysis [12, 21, 40]. Different approaches require the addition of specific information, such as cost, timing or probability of occurrence, to evaluate a model [22, 24, 40, 67]. Demanding formal, quantitative or detailed representations for high-level social concepts helps the flexibility and usability of these approaches for early RE.

4.2.6 Involvement of stakeholders

An active part of the early RE process is important to provide and validate domain information for stakeholders. In general, current procedures do not focus on the role of the stakeholder in the analysis. At the beginning of the process, most procedures take input as the first step to frame the query over the model without the modeller and provide analysis results at the end of the output. Some procedures allow for expert intervention at certain points [7, 19]. Typically, in these procedures, the participation of “experts” is seen as a necessary step to enhance the model or analysis with domain knowledge, but it is not explicitly encouraged as a means for engaging the user. In the goal model analysis, encouraging user involvement in a structured and clear way allows a higher level of user input, encouraging iteration over the correctness and completeness of the model, and increases the chances of obtaining stakeholder consensus in the new system.

4.2.7 Analysis and reasoning on modelling

Existing procedures support a wide range of analysis questions over models. However, most procedures require the addition of specific quantitative or qualitative or automatic production of results on high-level models. Procedures provide a wide range of analysis capabilities to keep information available from the user’s role for early RE models. Hence, the analysis results may be conflicting, incomplete or inaccurate to a certain degree, and should receive appropriate weight in domain understanding and decision making as part of a methodology for early RE exploration.

4.2.7.1 Qualitative analysis

In this section, some of the previous qualitative goal analysis methods with their drawbacks are listed.

Van Lamsweerde [67] introduces a qualitative reasoning model for the alternative estimation. Initially, the alternative solution for the contribution to different softgoals is assessed qualitatively. Next, these contributions are propagated toward the upper level, and the softgoal graphs are marked as “ $+$ ” or “ $++$ ” based on the strength of the positive contribution; whereas the negative contributions in terms of conflict links are termed as “ $-$ ” or “ $--$ ”. The propagation is considered recursively till the softgoals at the top level accept a single label. The limitations of this model are listed below.

•
The propagation rule converts the labels into “inconclusive”.
•
According to the system-specific phenomena, the link weight and labels have no clear meaning.
•
The evaluation of goals is offered roughly.

Concerning these problems, a lightweight quantitative alternative validation system is proposed by Lamsweerde [66], which combines the goals and softgoals into the KAOS model. The alternatives for contributions to all leaf nodes are assessed using quantitative estimations. The relative importance of the leaf softgoals are assigned with different weights. For an overall comparison, the scores and weight matrix are collected in the weighted matrix. For each softgoal, this model uses the variables such as ideal target value, maximum acceptable value and gauge variable. These variables are determined based on the system specifications. Hence, to design a goal model, the system specifications have to be analysed clearly. Some of the limitations of this model cause difficulty in applying for the higher complexity and the large systems, and also when the same label is received by two or more goals, then ambiguity arises in the decision-making process.

Mylopoulos et al. [49] introduce a business-goals-based goal-oriented analysis technique to explore a design alternative to estimate the feasibility and desirability of the system. The goal structure is represented by AND-OR decomposition and includes five steps below.

1.
Analysis of the goal
2.
Softgoal analysis
3.
Analysis of softgoal correlation
4.
Analysis of goal correlation
5.
Alternative estimation.

A case study is provided for the evaluation of this approach explained with the meeting scheduler. From the above four steps, the softgoal and goal decomposition has been constructed, and the evaluation of goal decomposition is performed in terms of softgoal hierarchy. Also, a selection of the set of softgoals is made for the evaluation that satisfies all given goals and the overall satisfaction for the top-level softgoals.

Giorgini et al. [24] present a goal model based on a qualitative formalisation and label propagation algorithm for the formal reasoning of goals. They design a model for a goal that incorporates qualitative relationships of goals and contradictory conditions. A goal relationships label with “ $+$ ”, “ $-$ ” is introduced that defines the positive and negative contributions in the satisfaction of other goals. The labelling of goals uses the semantic representation of the new goal. This is implemented with a Java platform with two different processes; labelling the propagation algorithm with a qualitative and quantitative formalisation. Also, two different sets of experiments are performed for qualitative label propagation and quantitative label propagation algorithms. A set of goals were assigned for the events, and some goals and their significance’s are eminent in the qualitative algorithm. After five iterations, a steady state is achieved in the qualitative task. On the other hand, the numeric weights are assigned for the relationships “ $+$ ”, “ $-$ ” and “ $-$ s” in the quantitative algorithm task. For each goal, final values are calculated using the propagation algorithm, and the outcome is ensured by the result predicted from the qualitative algorithm. Also, a numerical approach is utilised for the accurate estimation related to the final values of goals/events. For new relationships, quantitative semantics is used, which is obtained from the probabilistic model and also needs mathematical understanding, because it uses first-order logic.

Horkoff and Yu [32, 31] designed goal and agent-oriented models for the qualitative analysis to solve the problem that arises during the initial stage of requirements engineering. Also, to understand the problem domain, an interactive evaluation procedure is introduced for the evaluation of alternatives that require the intervention of a customer [35, 23]. Here, the alternatives are termed as process design or a system, capabilities, courses of actions and commitments. For the manual analysis, an informal process based on the $i^{}$ framework has been introduced. This analysis is performed based on raising the question, “How alternative design option is an effective one with respect to model goals?” A satisfaction or denial degree is assigned for the intentional elements based on a set of qualitative labels. These labels are propagated via model links using a propagation algorithm. When multiple or partial goals become conflicted, then a human judgment is performed for the analysis of satisfaction or denial of a softgoal. However, the final satisfaction and denial scores for each actor are analysed based on the original queries. The result estimates whether a choice of design is satisfied or not, and then the further analysis and refinement processes are performed. Experimental analyses are evaluated with various case studies to estimate where this process can be applied. The results prove that the procedure could be applied for a better understanding of the model and domain. However, during the decision-making process, ambiguity arises when multiple goals are attained with the same label, and it suffers from validity issues due to a minimum number of participants.

4.2.7.2 Quantitative analysis

Several existing models related to quantitative goal analysis approaches and their drawbacks are explained below.

Letier and Van Lamsweerde [40] present a heavy-weight model from the number interpretations related to probability. Lamsweerde et al. [66] introduce an approach for the goal satisfaction estimation in terms of partial degree, and computed the impact of alternatives for higher-level goals in terms of partial satisfaction. The alternatives of goal satisfaction degrees are evaluated based on qualitative and quantitative reasoning approaches. Softgoal predictions are also made by Bayesian networks techniques. For partial goal satisfaction estimation precision, an objective function and quality variables are utilised. For evaluating alternative design, the quality variables and the objective function are specified using the five heuristics model. The accurate specification of objective functions is specified by the probabilistic extension of temporal logic. For each alternative, the objective function values are computed by propagation rules that combine the quality variables of subgoals to the parent goals. However, the actual estimation of the objective function is performed via ad-hoc use of mathematical software. For the complex equations, the estimation leads to difficulty, and hence dedicated tools are provided to handle such computations effectively. Further, this model is improved to overcome the parameter uncertainties on calculations using confidence intervals.

A hybrid approach by the integration of qualitative and quantitative techniques was designed by Amyot et al. [5] that performs GRL analysis and calculates the satisfaction degree for the intentional elements and actors. The subgroups of intentional elements include the satisfaction scores, and these values are propagated based on propagation algorithm via contribution, dependency links and decomposition to new intentional elements. For the same GRL model, this process performs at several intervals by assigning various intentional strategies with a different subset of intentional elements. Two evaluation procedures, namely qualitative and quantitative, with the range of integer values lies within [ $-$ 100 … 100] analysis are selected. Initially, for other intentional elements, these two attribute scores are set to none and 0, respectively. In URN models, an Eclipse-based editor, the jUCMNav tool is employed for the implementation of three algorithms such as qualitative, quantitative and hybrid evaluation. The evaluation of mixed propagation performs from intentional elements, which are neither roots nor leaves. However, this model does not design a better goal model, and also, assigning exact numeric values to requirements is challenging.

Liaskos et al. [41] present a model for the prioritisation and the indication of preference requirements. This model distinguishes the mandatory goals from preference goals and also suggests a technique for finding other ways to achieve the mandatory and the preference goals. In preference goals, an optimised preference function is obtained by assigning an analytic hierarchy process (AHP) weights. However, this model cannot be applied to the larger analysis.

Franch [17] proposes an agent-oriented model designed related to the quantitative aspect which is constructed based on the $i^{}$ language. The conceptual model of this language is expressed based on unified modelling language (UML) and object constraint language (OCL). The structural indicators are adopted to obtain the structural metrics that compute the $i^{*}$ model properties include dependencies, actors and other factors. For the selection of alternatives during decision making, this method requires expert judgment. Also, this model is completely quantitative, because accurate data is obtained based on some degree of qualitative reasoning techniques.

To solve the issues in the decision of NFR analysis, Mairisa et al. [44] discusses a multi-criteria decision analysis (MCDA) that performs the analysis and evaluation of alternative design solutions. This model found the best design and satisfies the conflicts on NFRs. Also, an ideal solution is determined to find the alternative based on a goal-based technique named technique for order of preference by similarity to ideal solution (TOPSIS). However, an evaluation analysis of this approach was not provided in this approach.

Goncalves and Krishna [28] introduce a quantitative-based model for operationalisation in the extended NFR. The suitable operationalisation is determined based on its preferences and the progressive range of values of its children within the extended model. Based on the consideration of complexity, time and space, an optimal path is suggested for the combination of operationalisation at any particular time. Here, a simulation of this model is carried out using a banking system case study. This work is extended by adding change management in agents [37]. Whenever any changes occur, an optimal decision path is determined to estimate the agents with the variations such as softgoal weight or change in contribution values. Based on probability criteria, an optimisation model is designed, and an experimental evaluation is carried out. The result shows that the model cannot be applied directly and requires changes to be made in the original model.

4.2.7.3 Optimisation in goal analysis

In real-time competitive circumstances, decision making encompasses goals, alternative options, actors, decision-makers and criteria. The main factor of decision making depends on finding all the ecological factors and computing based on its objectives. The ideal solution is discovered by the decision-makers in which the optimisation methods determine the optimal alternatives from the list of possible options based on techniques such as linear, quadratic and non-linear programming [52].

Some of the existing optimisation methods and their limitations in the goal analysis task are listed below:

To determine the alternative design option, Heaven and Letier [30] present an extended model from their earlier work by utilising a multi-objective structure from the KAOS model. In their existing goal models analysis, a heuristics set and formal semantics are utilised. However, this model is not applicable for a greater number of design alternatives and does not automate the model analysis. Hence, an automated technique is designed by overcoming these limitations, and identifies the optimal alternatives among them. For the simulation, a stochastic simulation is employed where the input uses the list of design samples and size. For the particular design, the simulation is performed by the probability distributions and quantitative goal model equations. Further, the satisfaction scores are obtained for the simulation process. The simulation is performed based on the MATLAB platform presented based on the London Ambulance Service goal model. However, there is no systematic technique for the computation of objective function.

For the NFR framework, Affleck et al. [2, 1] present a linear programming optimisation that focuses on operationalisation minimisation. To support the decision-making process, an improved version of the quantitative model from their previous work ([3]) is utilised. Here, the weights are employed to the leaf softgoals and to the links between operationalisation and softgoals. The calculation of the leaf softgoal scores, operational scores, actual attainment and optimal scores are computed using these predicted weight values. Also, the linear programming method is applied for the objective function estimation concerning decision variables. To compute the result, objective functions, variables and constraints are applied as the input to the linear solver. The main objective of this work is to generate an optimal set of operationalisation represented in terms of minimisation or maximisation problem. In the extended goal graph, the propagation procedure from the leaf to the root softgoals is added. The implementation process is carried out using LPSolve, and the outcome shows that the process is better when there is a greater set of relationships between operationalisation and softgoals. For minimum operation, a single-objective optimisation is utilised that increases the overall satisfaction score of the NFR.

The quantitative input values are determined by sensitivity analysis. If the potential value goes beyond the given range, then the required action is performed accordingly. However, the values allowed to leaf softgoals are subjective, and assigning accurate value to leaf goals is difficult.
4.2.8 Usability and selection of decision-making methodologies for goal analysis

The analysis results can be used as a form of the rationale for decisions. However, users must be able to easily compare results to understand why it was believed that a particular analysis scenario was preferable to another. If procedures move away from full automation, users need to use their domain knowledge; decisions should be captured and somehow stored on the model. Modellers should be able to return these decisions to remember why they were made and change them if desired. Current interactive goal model analysis procedures do not provide support for organisation on model decisions. It would be particularly useful to allow modellers to capture the free-form rationale for the decision and attach it in some way over the model. Although the work of Maiden et al. [43] supports storage and management of satisfaction arguments over model decisions, their approach applies these arguments to limited structures and does not emphasise modification to arguments. Hence, existing techniques often review related procedures that aim to specifically guide the selection of techniques for goal analysis based on the requirements or characteristics of the goal model domain.

5. Conclusion and future works

In this paper, the various efforts of GORE research has been summarized by specifying the importance of requirements engineering, Goal oriented requirements engineering and GORE methods. From the review of the goal analysis literature and its limitations, we find that there is a need for methods to deal with linguistic terms of requirements. This paper aims to fill the research gaps by using fuzzy numbers for the linguistic representation of stakeholders’ requirements. Furthermore, numerous conflicting and competing objective functions encounter real-world business issues in simultaneous optimisation. For goal analysis procedures, several existing approaches are unable to address these issues examined by the literature. To fill the research gaps, efficient methods are required to eliminate subjective preference and manage vague requirements based on goal programming and multi-objective optimisation. Also, several approaches are needed based on analytical decision-making techniques to address requirements with opposing objective functions.

Footnotes

Author’s Bios

Dr. Sreenithya Sumesh has a Doctor of Philosophy in Computer Science from Curtin University, Australia. Her current research interests include software engineering, requirements engineering, agent modelling, machine learning, data mining and smart grids. She has presented and published many scientific research publications in reputed journals and conferences.

Associate Professor Dr. Aneesh Krishna is currently acting Discipline Lead of Computing within the School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Australia. He holds a PhD in computer science from the University of Wollongong, Australia. He was a lecturer at the School of Computer Science and Software Engineering, University of Wollongong, Australia (from February 2006 – June 2009). He has expertise in large-scale complex software development, model-driven development and evolution, service computing, software quality, requirements engineering, agent systems, formal methods, data mining, computer vision, and machine learning. His research is (or has been) funded by the Australian Research Council (ARC), Australian government agencies (Department of Defence, Australian Academy of Science, NSW State Emergency Service, Department of Industry, Science, Energy and Resources) as well as various industry partners. He has widely published (with more than 130 articles) in the above areas in top journals and international conferences. He has been on the organising committee,

served as invited technical program committee member of many conferences and workshops in the areas related to his research. He has supervised 9 PhD and 2 MPhil students to the successful completion of their degrees. He works closely with industry and undertakes practical; real-world industry-focussed research projects. He has worked (and led in most instances) on several successful projects with companies like Woodside Energy, Amristar Solutions, Thales, Deloitte, IBM, Immersive Technologies, Gaia Resources, AFG Group, Autism West Incorporated, BW Solar Australia, Western Australia Dementia Training Center and Andrew Corporation.

References

Affleck

and Krishna

, Supporting quantitative reasoning of non-functional requirements: A process-oriented approach, in: Proceedings of the International Conference on Software and System Process, Zurich, Switzerland, June 2–3, 2012, IEEE Press, Switzerland, 2012, pp. 88–92.

Affleck

Krishna

and Achuthan

, Optimal selection of operationalizations for non-functional requirements, in: Proceedings of the Ninth Asia-Pacific Conference on Conceptual Modelling-Adelaide, Australia, January 29–Feburary 1, 2013, Vol. 143, Australian Computer Society, Inc., 2013, pp. 69–78.

Affleck

Krishna

and Achuthan

, Non-functional requirements framework: A mathematical programming approach, The Computer Journal 58(5) (2014), 1122–1139.

Affleck

Krishna

and Achuthan

N.R.

, Non-functional requirements framework: A mathematical programming approach, The Computer Journal 58(5) (2015), 1122–1139.

Amyot

Ghanavati

Horkoff

Mussbacher

Peyton

and Yu

, Evaluating goal models within the goal-oriented requirement language, International Journal of Intelligent Systems 25(8) (2010), 841–877.

Ashamalla

Beydoun

and Low

, Model driven approach for real-time requirement analysis of multi-agent systems, Computer Languages, Systems & Structures 50 (2017), 127–139.

Asnar

and Giorgini

, Analyzing business continuity through a multi-layers model, in: International Conference on Business Process Management, Springer, 2008, pp. 212–227.

Ayala Martínez

C.P.

Cares

Carvallo Vega

J.P.

Grau Colom

Haya

Salazar

Franch Gutiérrez

Mayol Sarroca

and Quer

, A comparative analisys of i*-based agent-oriented modeling languages, in: SEKE 2005, the 17th International Conference on Software Engineering & Knowledge Engineering: Technical Program: July 14–16, 2005, Taipei, Taiwan, Republic of China: Proceedings, 2005, pp. 43–50.

Beydoun

Krishna

Ghose

and Low

G.C.

, Towards ontology-based mas methodologies: ontology-based early requirements, in: Information Systems Development, Springer, 2009, pp. 923–935.

10.

Boardman

A.E.

Greenberg

D.H.

Vining

A.R.

and Weimer

D.L.

, Cost Benefit analysis: concepts and practice, Cambridge University Press, 2017.

11.

Bresciani

Perini

Giorgini

Giunchiglia

and Mylopoulos

, Tropos: An agent-oriented software development methodology, Autonomous Agents and Multi-Agent Systems 8(3) (2004), 203–236.

12.

Bryl

Giorgini

and Mylopoulos

, Designing cooperative is: Exploring and evaluating alternatives, in: OTM Confederated International Conferences “On the Move to Meaningful Internet Systems”, Springer, 2006, pp. 533–550.

13.

Chawla

, Goal oriented requirements engineering for blockchain based food supply chain, International Journal of Software Engineering and Computer Systems 6 (2020), 94–103.

14.

Chung

Nixon

B.A.

and Mylopoulos

, Non-functional requirements in software engineering, Vol. 5, Springer Science and Business Media, 2012.

15.

Dardenne

Fickas

and van Lamsweerde

, Goal-directed concept acquisition in requirements elicitation, in: Proceedings of the 6th International Workshop on Software Specification and Design, Como, Italy, IEEE Computer Society Press, Italy, 25–26 Oct. 1991, pp. 14–21.

16.

Dybå

Prikladnicki

Rönkkö

Seaman

and Sillito

, Qualitative research in software engineering, Empirical Software Engineering 16(4) (2011), 425–429.

17.

Franch

, On the quantitative analysis of agent-oriented models, in: CAiSE, Springer Verlag Berlin, Heidelberg, Vol. 6, 2006, pp. 495–509.

18.

Franch

Grau

and Quer

, A framework for the definition of metrics for actor-dependency models, in: Proceedings. 12th IEEE International Requirements Engineering Conference, 2004, IEEE, 2004, pp. 348–349.

19.

Franch

López

Cares

and Colomer

, The i* framework for goal-oriented modeling, in: Domain-Specific Conceptual Modeling, Springer, Switzerland, 2016.

20.

Fuxman

Liu

Mylopoulos

Pistore

Roveri

and Traverso

, Specifying and analyzing early requirements in tropos, Requirements Engineering 9(2) (2004), 132–150.

21.

Fuxman

Pistore

Mylopoulos

and Traverso

, Model checking early requirements specifications in tropos, in: Proceedings Fifth IEEE International Symposium on Requirements Engineering, IEEE, 2001, pp. 174–181.

22.

Gans

Jarke

Kethers

and Lakemeyer

, Continuous requirements management for organisation networks: A (dis) trust-based approach, Requirements Engineering 8(1) (2003), 4–22.

23.

Gaol

F.L.

Danny

and Matsuo

, Application of organization goal-oriented requirement engineering (ogore) methods in erp-based company business processes, Open Engineering 9(1) (2019), 545–553.

24.

Giorgini

Mylopoulos

Nicchiarelli

and Sebastiani

, Reasoning with goal models, ER 2 (2002), 167–181.

25.

Giorgini

Nicchiarelli

Mylopoulos

et al., Formal reasoning with goal models, in: Proc. of Intâ€™l Conference of Conceptual Modeling. Tampere, Finland, October 7–11, 2002, Lecture Notes in Computer Science volume 2503, Springer Finland, 2002, pp. 167–181.

26.

Goncalves

and Krishna

, Dynamic non-functional requirements based model-driven agent development, in: 2015 24th Australasian Software Engineering Conference, IEEE, 2015, pp. 128–137.

27.

Goncalves

and Krishna

, Incorporating change management within dynamic requirements-based model-driven agent development, The Computer Journal 60(7) (2017), 1044–1077.

28.

Goncalves

J.Z.

and Krishna

, Optimal requirements-dependent model-driven agent development, in: Transforming Healthcare Through Information Systems, Springer, 2016, pp. 135–151.

29.

Grau

Franch

and Ávila

, J-prim: A java tool for a process reengineering i* methodology, in: 14th IEEE International Requirements Engineering Conference (RE’06), IEEE, 2006, pp. 359–360.

30.

Heaven

and Letier

, Simulating and optimising design decisions in quantitative goal models, in: 2011 IEEE 19th International Requirements Engineering Conference, IEEE, 2011, pp. 79–88.

31.

Horkoff

Aydemir

F.B.

Cardoso

Maté

Paja

Salnitri

Piras

Mylopoulos

and Giorgini

, Goal-oriented requirements engineering: An extended systematic mapping study, Requirements Engineering 24(2) (2019), 133–160.

32.

Horkoff

and Yu

, Evaluating goal achievement in enterprise modeling – an interactive procedure and experiences, in: IFIP Working Conference on The Practice of Enterprise Modeling, Springer, 2009, pp. 145–160.

33.

Kaiya

Horai

and Saeki

, Agora: Attributed goal-oriented requirements analysis method, in: Requirements Engineering Proceedings. on IEEE Joint International Conference Germany, IEEE, Germany, 4–6 Aug. 2002, pp. 13–22.

34.

Kasauli

Knauss

Horkoff

Liebel

and de Oliveira Neto

F.G.

, Requirements engineering challenges and practices in large-scale agile system development, Journal of Systems and Software 172 (2021), 110851.

35.

Kavakli

, Goal-oriented requirements engineering: A unifying framework, Requirements Engineering 6(4) (2002), 237–251.

36.

Kelly

, Decision Making Using Game Theory: An Introduction for Managers, Cambridge University Press, Cambridge, United Kingdom, isbn: 9780521814621, Cambridge University Press, Cambridge, United Kingdom, ISBN: 9780521814621, 03 2003.

37.

Krishna

Goncalves

and Krishna

, Optimal requirements-dependent model-driven agent development, transforming healthcare through information systems, in: Transforming Healthcare Through Information Systems, Springer International Publishing, 2016, pp. 135–151.

38.

Krishna

Vilkomir

S.A.

and Ghose

A.K.

, Consistency preserving co-evolution of formal specifications and agent-oriented conceptual models, Information and Software Technology 51(2) (2009), 478–496.

39.

Lázaro

and Marcos

, An approach to the integration of qualitative and quantitative research methods in software engineering research, PhiSE (2006), 240.

40.

Letier

and Van Lamsweerde

, Reasoning about partial goal satisfaction for requirements and design engineering, in: Proceedings of the 12th ACM SIGSOFT Twelfth International Symposium on Foundations of Software Engineering, 2004, pp. 53–62.

41.

Liaskos

McIlraith

S.A.

Sohrabi

and Mylopoulos

, Integrating preferences into goal models for requirements engineering, in: Requirements Engineering Conference (RE), 2010 18th IEEE International, New South Wales, Australia, September 27–October 1, 2010, IEEE Computer Society USA, 2010, pp. 135–144.

42.

Liaskos

McIlraith

S.A.

Sohrabi

and Mylopoulos

, Representing and reasoning about preferences in requirements engineering, Requirements Engineering 16(3) (2011), 227–249.

43.

Maiden

Lockerbie

Randall

Jones

and Bush

, Using satisfaction arguments to enhance i* modelling of an air traffic management system, in: 15th IEEE International Requirements Engineering Conference (RE 2007), IEEE, 2007, pp. 49–52.

44.

Mairiza

Zowghi

and Gervasi

, Utilizing topsis: A multi criteria decision analysis technique for non-functional requirements conflicts, in: Requirements Engineering, Springer, 2014, pp. 31–44.

45.

Miller

Sterling

Beydoun

and Taveter

, Requirements elicitation and specification using the agent paradigm: The case study of an aircraft turnaround simulator, IEEE Transactions on Software Engineering 40(10) (2014), 1007–1024.

46.

Morandini

Nguyen

D.C.

Perini

Siena

and Susi

, Tool-supported development with tropos: The conference management system case study, in: International Workshop on Agent-Oriented Software Engineering, Springer, 2007, pp. 182–196.

47.

Mylopoulos

Chung

Liao

Wang

and Yu

, Exploring alternatives during requirements analysis, IEEE Software 18(1) (2001), 92–96.

48.

Mylopoulos

Chung

and Nixon

, Representing and using nonfunctional requirements: A process-oriented approach, IEEE Transactions on Software Engineering 18(6) (1992), 483–497.

49.

Mylopoulos

Chung

and Yu

, From object-oriented to goal-oriented requirements analysis, Communications of the ACM 42(1) (1999), 31–37.

50.

Neumann

P.J.

and Sanders

G.D.

, Cost effectiveness analysis 2.0, N Engl J Med 376(3) (2017), 203–205.

51.

Oliveira

L.S.d.

and Saramago

S.F.

, Multiobjective optimization techniques applied to engineering problems, Journal of the Brazilian Society of Mechanical Sciences and Engineering 32(1) (2010), 94–105.

52.

Ponsard

and Darimont

, Towards multi-objective optimisation of quantitative goal models using constraint programming, in: ICORES, 2020, pp. 286–292.

53.

Robinson

, Cost effectiveness analysis, British Medical Journal 307(6907) (1993), 793–795.

54.

Saaty

R.W.

, The analytic hierarchy process – what it is and how it is used, Mathematical Modelling 9(3–5) (1987), 161–176.

55.

Sadiq

and Jain

, Applying fuzzy preference relation for requirements prioritization in goal oriented requirements elicitation process, International Journal of System Assurance Engineering and Management 5(4) (2014), 711–723.

56.

Shaked

and Reich

, Requirements for model-based development process design and compliance of standardized models, Systems 9(1) (2021), 3.

57.

Shofi

I.M.

Nugraha

Anggraini

and Hakiem

, Implementation of organization goal-oriented requirements engineering (ogore) method in designing a muhammadiyah high school library information system, in: 2020 3rd International Conference on Computer and Informatics Engineering (IC2IE), IEEE, 2020, pp. 366–371.

58.

Subramanian

Krishna

and Kaur

, Optimal reasoning of goals in the i* framework, in: Software Engineering Conference (APSEC), New Delhi India, 1–4 Dec, APSEC Asia-Pacific, 2015, pp. 346–353.

59.

Subramanian

Krishna

and Kaur

, Reasoning about goal satisfaction for early requirements engineering in the i* framework using inter-actor dependency, in: 19th Pacific Asia Conference on Information Systems, Singapore, July 5–9, PACIS 2015, Singapore, 2015, p. 89.

60.

Subramanian

Krishna

and Kaur

, Optimal goal programming of softgoals in goal-oriented requirements engineering, in: 20th Pacific Asia Conference on Information Systems, PACIS 2016, Chiayi, Taiwan, June 27–July 1, 2016, PACIS, Taiwan, 2016, pp. 202–217.

61.

Subramanian

Krishna

Kaur

and Gopalan

, Quantitative reasoning of goal satisfaction in the i* framework, in: The 27th International Conference on Software Engineering and Knowledge Engineering, SEKE 2015, PA, USA, July 6–8, 2015, KSI Research Inc. and Knowledge Systems Institute Graduate School USA, 2015, pp. 666–669.

62.

Subramanian

C.M.

Krishna

and Kaur

, Game theory-based requirements analysis in the i* framework, The Computer Journal 61(3) (2018), 427–446.

63.

Sumesh

Krishna

and Subramanian

C.M.

, CEA based reasoning with the i* framework, in: 22nd Pacific Asia Conference on Information Systems, PACIS 2018, Yokohama, Japan, June 26–30, 2018, Association for Information Systems, 2018, p. 174.

64.

Sumesh

Krishna

and Subramanian

C.M.

, Game theory-based reasoning of opposing non-functional requirements using inter-actor dependencies, The Computer Journal 62(11) (2019), 1557–1583.

65.

Van Lamsweerde

, Goal-oriented requirements engineering: A guided tour, in: Proceedings Fifth Ieee International Symposium on Requirements Engineering, IEEE, 2001, pp. 249–262.

66.

van Lamsweerde

, Goal-oriented requirements enginering: A roundtrip from research to practice, in: 12th IEEE International Requirements Engineering Conference Proceedings, Kyoto Japan, 6–10 September, IEEE Computer Society, Japan, 2004, pp. 4–7.

67.

van Lamsweerde

, Reasoning about alternative requirements options, in: Conceptual Modeling: Foundations and Applications, Springer, 2009, pp. 380–397.

68.

Vilkomir

S.A.

Ghose

A.K.

and Krishna

, Combining agent-oriented conceptual modelling with formal methods, in: 2004 Australian Software Engineering Conference. Proceedings, IEEE, 2004, pp. 147–155.

69.

Wang

and Lespérance

, Agent-oriented requirements engineering using congolog and i*, in: Agent-Oriented Information Systems Workshop (AOIS-2001). Montreal, Canada, 2001, pp. 59–78.

70.

Wang

X.-T.

and Xiong

, An integrated linguistic-based group decision-making approach for quality function deployment, Expert Systems with Applications 38(12) (2011), 14428–14438.

71.

Wang

Zhou

and Du

, Toward practical adoption of i* framework: An automatic two-level layout approach, Requirements Engineering 20 (2021), 1–23.

72.

Waters

Padgham

and Sardina

, Improving domain-independent intention selection in bdi systems, Autonomous Agents and Multi-Agent Systems 29(4) (2015), 683–717.

73.

Yazdania

Linb

Y.-T.

Caic

and Huanga

, A game theory perspective on requirement-based engineering design, in: Disciplinary Convergence: Implications for Systems Engineering Research, 15th Annual Conference on System Engineering Research, Redondo Beach, CA, System Engineering Research, CA, 23–25 Mar. 2017, pp. 901–910.

74.

, Agent-oriented modelling software versus the world, in: International Workshop on Agent-Oriented Software Engineering, Montreal, Canada, Springer, Berlin, Heidelberg, Vol. 2222, 29 May 2001, pp. 206–225.

75.

, Modeling strategic relationships for process reengineering, Social Modeling for Requirements Engineering 11(2011) (2011), 66–87.

Challenges and review of goal-oriented requirements engineering based competitive non-functional requirements analysis

Abstract

Keywords

1. Introduction

2. Research issues related to competitive non-functional requirements analysis

3. Challenges in agent-based goal model reasoning for early requirements engineering

Table 1 Analysis methods in GORE

3.2 Completeness of modelling

(a) knowing where to stop modelling (b) knowing if the amount of information collected is sufficient to support useful analysis, reasoning and understanding of the domain. 3.3 Accuracy of model analysis

3.4 Understanding of the domain

3.5 Flexibility in modelling

3.6 Involvement of stakeholders

3.7 Analysis and reasoning on modelling

3.8 Usability and selection of decision-making methodologies for goal analysis

4.1 Conceptual foundation

4.2 Identification of research gaps in goal model reasoning

4.2.1 Complexity in model analysis

4.2.2 Completeness of modelling

4.2.3 Accuracy of model analysis

4.2.4 Understanding of the domain

4.2.5 Flexibility in modelling

4.2.6 Involvement of stakeholders

4.2.7 Analysis and reasoning on modelling

4.2.7.1 Qualitative analysis

4.2.7.2 Quantitative analysis

4.2.7.3 Optimisation in goal analysis

5. Conclusion and future works

Footnotes

Author’s Bios

References

Table 1
Analysis methods in GORE

(a)
knowing where to stop modelling
(b)
knowing if the amount of information collected is sufficient to support useful analysis, reasoning and understanding of the domain.

3.3 Accuracy of model analysis