Journal of Intelligent & Fuzzy Systems: Applications in Engineering and Technology

Clustering analysis quantifies similarities (dissimilarities) between objects in a given dataset and discovers the hidden characteristics of each cluster. However, researchers often have difficulty in setting optimal parameters for clustering analysis when they attempt to obtain the optimal clustering. This work presents an entropy-based efficient clustering technique utilizing principles of genetic algorithm (GA), unlike previous clustering method [24] which employs parameter setting. The proposed method considers the data spread to determine the adaptive threshold within parameters optimized by genetic algorithm. The fitness function of genetic algorithm is defined as clustering accuracy. Four datasets in the UCI database are selected as the experimental data to compare the accuracy of the proposed algorithm with the three clustering methods. Results of this study demonstrate that the proposed algorithm outperforms listing methods.

It is shown in this paper that a qualitative mapping model of pattern recognition can be induced by the basic princeple that the pattern P({(x $_{j}$ , y'(x $_{j}$ )}) constructed by the set of the pair of varibale x $_{j}$ and its computing value y'(x $_{j}$ ) of a function y = f(x) could be considered as the graph of y(x), and a logic circuit, called Attribute Network Computing can be constructed by the relation between the input and output of qualitative mapping, such that not only a lot of artificial intelligent methods, such as Expert Syatem, Artificial Neural Network, and Support Vector Machine can be fused and unified in it, but also a lot of pattern recognition can be implemented by it.

Process capability indices (PCIs) are mainly used in industry in order to measure the capability of a process to produce products meeting specifications. Traditionally, the specifications are defined as crisp numbers. Sometimes, specification limits can be expressed as linguistic terms. Traditional PCIs can not be applied for this situation. In this paper, PCIs are analyzed under fuzzy environment. The fuzzy process capability analyses are developed when the specifications limits are represented by triangular and trapezoidal fuzzy numbers. The developed approaches are applied to teaching processes and the fuzzy PCIs are compared based on two different fuzzy ranking methods.

This paper reports a real-time approach for backing-up car control, in which an image sensor and a fuzzy logic controller are utilized. In the system, a CCD camera is employed as an image sensor for detecting motions of a backing-up car. The fuzzy logic controller receives the motion information from the image sensor, and then generates the steering signal of the car to complete a closed-loop control system of backing-up an autonomous vehicle. The MATLAB/Simulink® are installed in a PC to execute the control action through a model car, driven by two electric motors. Both simulations and experimental results are presented to verify the feasibility of the proposed method.

Quality function deployment (QFD) is a product/service design and improvement tool used to achieve higher customer satisfaction. Basically, QFD is a transformation of vague and imprecise customer needs into measurable product/service attributes. This article integrates compromise programming based goal programming into the QFD process to determine to what extent the product/service attributes should be improved. The fuzzy set theory is applied to the model to deal with the imprecise nature of data. Differing from existing QFD applications, our proposed methodology applies analytic network process (ANP) to evaluate the inner dependencies among customer needs, among product attributes and also the relationships between them. Furthermore, it determines the best product/service in the market as the goal employing compromise programming. Finally, the methodology ends with the goal programming method which consists of this predefined goal and the product/service provider's budget limitation. The aim is to improve the performance values of the selected product/service according to the best one. A real-world application on e-learning products provided by the higher education institutions in Turkey illustrates the applicability of our proposed methodology.

In a bilevel decision problem, both the leader and the follower may have multiple objectives to optimize under certain constraints. In the meantime, these objective functions and constraints may contain some uncertain parameters. In addition, there may have multiple followers involved in a bilevel decision problem. These followers may share their individual decision variables with each other but keep individual objectives in reacting any of the leader's decisions, which is a common situation in real bilevel decision activities. This study deals with all above three issues, fuzzy parameters, multi-objectives, and multi-followers in a partial cooperative situation, at the same time. After a set of models for describing different cases of the fuzzy multi-objective multi-follower bilevel programming with partial cooperation (FMMBP-PC) problem, this paper develops an approximation branch-and-bound algorithm to solve this problem.

The crucial issue in many classification applications is how to achieve the best possible classifier with a limited number of labeled data for training. Training data selection is one method which addresses this issue by selecting the most informative data for training. In this work, we propose three data selection mechanisms based on fuzzy clustering method: center-based selection, border-based selection and hybrid selection. Center-based selection selects the samples with high degree of membership in each cluster as training data. Border-based selection selects the samples around the border between clusters. Hybrid selection is the combination of center-based selection and border-based selection. Compared with existing work, our methods do not require much computational effort. Moreover, they are independent with respect to the supervised learning algorithms and initial labeled data. We use fuzzy c-means to implement our data selection mechanisms. The effects of them are empirically studied on a set of UCI data sets. Experimental results indicate that, compared with random selection, hybrid selection can effectively enhance the learning performance in all the data sets, center-based selection shows better performance in certain data sets, border-based selection does not show significant improvement.

In the framework of the linguistic truth-valued logic, a linguistic truth-valued reasoning approach for decision making with both comparable and incomparable truth values is proposed in this paper. By using the lattice implication algebra, an 18-element linguistic truth lattice-valued logic system with linguistic hedges is established for the linguistic truth-valued logic to better express both comparable and incomparable truth values. Mathematical properties of disjunction, conjunction, negation and implication for the linguistic truth-valued propositional logic are further investigated respectively. As reasoning and operation are directly acted by linguistic truth values in the decision process, the issue on how to obtain the weight for rational decision making results is discussed. An illustration example shows the proposed approach seems more effective for decision making under a fuzzy environment with both comparable and incomparable linguistic truth values.

An effective algorithm capable of solving the multi-mode resource-constrained project scheduling problem (MRCPSP) is an essential component for project planning and control since it can fully exploit the available resources and minimize the makespan of a given project. The MRCPSP is extremely complex and is known to be NP-hard in the strong sense. On the basis of the principles of ant colony optimization (ACO), we therefore propose a constructive-oriented iterative algorithm to acquire satisfactory solutions of the MRCPSP within a reasonable amount of computation time. The proposed algorithm, namely ACO-MRCPSP, attempts to identify a project schedule with minimum completion time without violating precedence and resource constraints. ACO-MRCPSP is characterized by its use of a self-adaptive parameter control strategy to guide artificial ants to effectively construct feasible solutions for the MRCPSP. The performance of the proposed algorithm is evaluated by comparing it against other existing metaheuristic implementations, such as simulated annealing (SA) and genetic algorithms (GAs), in terms of overall completion time for a set of project instances obtained form the Project Scheduling Library (PSPLIB). Experimental results indicate that ACO-MRCPSP is a significant improvement compared with the previous attempts at solving the MRCPSP.

Real options valuation is a financial technique for evaluating investments under conditions of uncertainty, particularly uncertainty associated with market variables such as future product demand or the future value of an asset. The real option value of the investment opportunity is what a value-maximizing firm would pay the right to undertake the investment project with its inherent decision points. This paper proposes a fuzzy multi-criteria R&D project selection methodology based on the hierarchical fuzzy TOPSIS, which includes a fuzzy real options valuation model.

In Statistical Process Control (SPC), both variable and attribute control charts are widely used techniques to monitor and evaluate the process with respect to the related quality characteristics. In variable control charts, quality characteristics are measured on a numerical scale, using two main parameters mean and standard deviation. When data are composed of individuals from a process, individual (X) control charts are used for mean and moving range (MR) control charts are used for standard deviation. Generally, the data in variable control charts are assumed to be composed of crisp values. But a measurement system including operator, gage, and environmental conditions may produce "uncertain" or "vague" data. In this case, the fuzzy set theory is an available tool for evaluating the vague data. In this study, fuzzy control limits for individual (X) and moving range (MR) control charts with α-cuts are constructed by using α-level fuzzy median transformation techniques. The real-world data are used and the process is evaluated by the developed fuzzy control charts.