Design of an international trade target supplier evaluation system based on CART algorithm

Abstract

International trade, as an important component of economic exchange between countries, is of great significance for the economic development of each country and international cooperation. In international trade, the selection and evaluation of suppliers has always been a key issue. To ensure the smooth progress of trade and the controllability of quality, it is necessary to establish a target supplier evaluation system. This article used the CART (Classification and Regression Tree) algorithm to help identify and analyze the impact of key factors on supplier evaluation and classify and evaluate suppliers. The international trade target supplier evaluation system based on the CART algorithm was also constructed, and its performance was tested in the experimental section and compared with the international trade target supplier evaluation system based on traditional algorithms. According to the experimental results, it can be concluded that both the traditional algorithm and the CART algorithm performed well in terms of application effectiveness and system user satisfaction. In terms of application effectiveness, the average score of traditional algorithms was 4.3, with a rating range of 3.8 to 4.9, while the average score of the CART algorithm was 4.6, with a rating range of 4.2 to 5.0. The satisfaction rating of system users on the CART algorithm was slightly higher than that of traditional algorithms, indicating that the CART algorithm has better application effectiveness and user satisfaction in the design of international trade target supplier evaluation systems. The design of an international trade target supplier evaluation system based on the CART algorithm can also help enterprises reduce trade risks and improve the stability and reliability of the supply chain. It has important practical significance and application value for further promoting the development of international trade.

Keywords

Supplier evaluation system CART algorithm international trade procurement efficiency supply chain

1. Introduction

With the continuous advancement of globalization and the increasing frequency of international trade, the selection and evaluation of suppliers have become crucial. When facing the selection of numerous suppliers, establishing a scientific and effective evaluation system can help enterprises better screen and evaluate suitable suppliers, thereby improving trade efficiency, reducing trade costs, and increasing profits. By designing an international trade target supplier evaluation system, the aim is to provide scientific and systematic evaluation methods for enterprises to choose suitable suppliers in international trade. The purpose of supplier evaluation is to help enterprises reduce procurement risks, improve procurement efficiency, and achieve sustainable development goals. This system comprehensively considers key indicators such as supplier quality, cost, delivery time, and service, as well as evaluation of supplier stability and reliability, to provide a scientific decision-making basis for enterprises, improve the overall procurement management level, and promote the healthy development of international trade.

Conventional approaches for evaluating suppliers frequently lack science and accuracy and are instead dependent on subjective judgment and experience. As information technology has improved and the big data era has arrived, an increasing number of academics and businesses are attempting to leverage machine learning and advanced data analysis tools to increase the efficacy of supplier assessment. One of the most popular decision tree algorithms for classification and regression issues is the CART (Classification and Regression Trees) algorithm, which has a straightforward structure and is simple to comprehend and apply. The CART algorithm may build a decision tree model and identify the critical elements influencing supplier assessment by learning from and analyzing previous data. Businesses can choose suppliers quickly and accurately because of the model’s ability to automatically classify and evaluate based on fresh supplier data. The selection of an appropriate supplier is crucial in international trading. By using the CART algorithm, a decision tree model can be created that will assist businesses in rapidly and precisely identifying the best suppliers by automatically identifying and assessing the different attributes of possible suppliers based on historical data and other pertinent information. vendors. The manual analysis and judgment that are frequently used in the traditional supplier evaluation procedure are not only ineffective but also prone to subjectivity. This procedure can be automated by the supplier assessment system built on the CART algorithm, significantly increasing decision-making accuracy and efficiency. With the use of past data analysis, the CART algorithm can forecast prospective suppliers’ performance, assisting businesses in mitigating risk.

The following components primarily demonstrate the design and research contribution of the CART-based international trade target supplier evaluation system:

1.
The efficiency and accuracy of evaluating suppliers have increased: old methods of evaluating suppliers are typically dependent on experience and intuition, which are very subjective and lack objectivity. Human factor interference can be minimized, an objective, accurate, and thorough evaluation of suppliers can be accomplished, and the evaluation’s efficiency and accuracy may be increased by implementing the CART algorithm.
2.
Achieve automation and intelligence in supplier evaluation: The CART-based supplier evaluation system can classify and evaluate a substantial amount of supplier data automatically, as well as automate the processing of the data by predetermined evaluation indicators and standards. This allows for the automation and intelligence of supplier evaluation to be realized. This lowers evaluation costs while simultaneously increasing assessment accuracy and efficiency.
3.
Encouraged the use and advancement of the CART algorithm in the context of international trade: Adding the CART algorithm to the target suppliers’ evaluation system for international trade has not only increased the CART algorithm’s field of application but also encouraged its use and advancement in the context of international trade. This will increase the commercial value for businesses and support the growth of innovation in the area of international trade as well as digital transformation.

In conclusion, the research and design of the CART algorithm-based international trade target supplier evaluation system has significant theoretical and practical value. It also gives businesses access to more precise, effective, and intelligent methods and tools for evaluating suppliers, all of which will support the industry’s innovation and digital transformation.
2. Related work

2.1 Supplier evaluation system

The target supplier evaluation system can evaluate the ability and reputation of suppliers through objective indicators and standards, thereby helping enterprises choose suitable suppliers to improve trade efficiency and reduce risks [1, 2]. Yazdani, M proposed a comprehensive decision-making model, including decision-making experiments and evaluation laboratories, best-worst methods, and improved evaluation based on the average solution distance method, to solve the problem of supplier selection in the public procurement system considering the sustainable development goals, aiming to select the best suppliers for leading decision makers [3]. Liou, James JH proposed a new hybrid green supplier evaluation model, which combines support vector machines and fuzzy best and worst methods to select the most suitable green suppliers. Taking the data of multinational electronic product manufacturers as an example, a case study was conducted to integrate performance and prioritize green suppliers [4]. Bai and Chunguang proposed a social sustainability attribute decision-making framework to evaluate and select socially sustainable suppliers, using gray multi-standard decision support tools to determine the weight of social sustainability attributes and rank suppliers [5]. Chatterjee and Prasenjit proposed a two-phase model that uses a fuzzy analysis hierarchical process and fuzzy technology to evaluate and select suppliers through similarity to the ideal solution and rank potential suppliers selected through expert evaluation. The proposed model realizes the possibility of additional savings by developing stronger cooperation with the best suppliers [6]. Amiri and Maghsoud proposed a new model with a triangular fuzzy approach to sustainable supplier selection in the supply chain. The proposed fuzzy model is based on the best-worst method and $\alpha$ -cut analysis, in which decision-makers can determine the $\alpha$ value between 0.1 and 0.9 based on the level of uncertainty [7]. The CART algorithm is a powerful machine-learning method that can process complex data sets and generate easy-to-understand decision trees. It is applied to the design of the international trade target supplier evaluation system, which can process various types of data and is flexible, allowing decision-makers to easily understand the decision-making process of supplier evaluation and improve the accuracy and efficiency of the evaluation.

2.2 CART algorithm

The CART algorithm is a common decision tree algorithm that has extensive application value in supplier evaluation. As a key link of project management, project performance evaluation, supported by data analysis and mining technology, implements scientific and effective project performance evaluation in the context of the continuous development of modern information technology such as big data and cloud computing, which helps enterprise managers to timely and accurately obtain project performance, find projects with low-performance levels as early as possible and formulate plans to improve project performance [8, 9]. Carrizosa and Emilio believe that classification and regression trees and their variants are ready-made methods in machine learning. They compare the nature of decision variables and the required constraints, as well as the proposed optimization algorithms, explaining how these powerful formulas enhance the flexibility of the tree model and are more suitable for incorporating ideal attributes such as cost sensitivity, interpretability, and fairness, as well as processing complex data such as functional data [10]. Gomes and Cristiano Mauro Assis compared the CART algorithm and the unbiased algorithm (CTREE), as well as their predictive power, and found that the CART algorithm produces a tree with better result prediction. This result shows that for large data sets called big data, the CART algorithm may provide better results than the CTREE algorithm [11]. Li, Xinru proposed meta-CART to identify the interaction between multiple moderators, which provides a user-friendly feature for meta-CART analysis in R. The software package can be suitable for fixed and random effect meta-analysis and can handle bisection, classification, ordinal, and continuous adjuster [12]. Cheng Pingyang, taking A Landscape Architecture Planning and Research Institute as the research object, analyzed the characteristics of A Institute’s project and the current situation of performance evaluation, and proposed a performance evaluation index system for planning and design projects starting from the input and output elements of the project. He introduced data envelopment analysis and machine learning algorithms to design a project performance evaluation method based on the CART model [13]. By applying the CART algorithm to build a decision tree model, companies can more effectively evaluate suppliers in international trade, ensure that the most suitable partners are selected, and optimize supply chain management. This not only helps to improve procurement efficiency and product quality, but also reduces costs, reduces risks, and enhances the company’s competitiveness and sustainability.

To ensure the smooth progress of trade and the controllability of quality, this article constructed an international trade target supplier evaluation system based on the CART algorithm and compared it with the evaluation system under traditional algorithms for experiments. The experimental results indicated that the design of an international trade target supplier evaluation system based on the CART algorithm can not only help enterprises reduce trade risks, and improve the stability and reliability of the supply chain but also have important practical significance and application value for promoting the development of international trade.

3. Data processing based on CART algorithm

The CART algorithm, also known as the classification and regression tree algorithm, is a commonly used decision tree algorithm [14]. It can divide the dataset into multiple independent subsets, and build a decision tree model on each subset. The CART algorithm iteratively selects the optimal features and segmentation points to construct a decision tree, thereby achieving sample classification and prediction. In the design of an international trade target supplier evaluation system, the use of the CART algorithm can help identify and analyze the impact of key factors on supplier evaluation [15, 16]. By building a decision tree model, suppliers can be effectively classified into high, medium, and low levels for targeted evaluation and management [17].

The test set is used to evaluate the constructed decision tree model, including calculation accuracy, recall rate, accuracy rate, and other indicators, to evaluate the performance and stability of the model. According to the evaluation model, a supplier evaluation system is designed and implemented. This system can automatically extract features from supplier data and use the CART algorithm to evaluate and rank suppliers. The evaluation results can be visualized to help decision-makers better understand the strengths and weaknesses of suppliers.

There are many standards used at feature nodes. This article mainly introduces entropy and the Gini coefficient.

Entropy represents the uncertainty of random variables, which is a basic concept in information theory and probability statistics. A is assumed to be a discrete random variable with a finite number of values, and its probability distribution is:

$\displaystyle{Q}({{A}={a}_{j}})={Q}_{J},{J}=1,2,3,\ldots{m}$ (1)

The entropy of random variable A is defined as:

$\displaystyle{F}({A})=-\mathop{\sum}\limits_{j}^{m}{Q}_{J}\log Q_{J}$ (2)

To make the above formula meaningful, it is defined as $0\log 0=0$ . The value of entropy depends on the distribution of A and is independent of the value of A. Entropy can be regarded as a function of the distribution.

$\displaystyle{F}({Q})=-\mathop{\sum}\limits_{j}^{m}{Q}_{J}\log Q_{J}$ (3)

The Lagrand multiplier method is used to prove that the entropy of uniform distribution is the largest, and the Lagrand method can write ${F}({Q})$ as:

$\displaystyle{F}({Q})=-\mathop{\sum}\limits_{j}^{m}{Q}_{J}\log Q_{J}+{\beta}% \left({\mathop{\sum}\limits_{j}^{m}{Q}_{J}-1}\right)$ (4)

${F}({Q})$ derives each ${Q}_{J}$ , resulting in:

$\displaystyle\frac{\vartheta Q}{\vartheta{Q}_{J}}=-\text{In }{Q}_{J}-1+{\beta}% =0,{J}=1,2,\ldots,{m}$ (5)

The Gini index is the standard used by CART to select the optimal features for classification problems. There are $R$ classes, and the probability of the sample belonging to the $r$ class is ${q}_{r}$ , and the Gini index is:

$\displaystyle{G}({q})=\mathop{\sum}\limits_{{r}=1}^{R}{q}_{r}({1-{q}_{r}})=1-% \mathop{\sum}\limits_{{r}=1}^{R}{q}_{r}^{2}$ (6)

Condition $\mathop{\sum}\limits_{{r}=1}^{R}{q}_{r}=1$ is met.

By describing the CART algorithm, it is possible to better understand and apply it to the design of an international trade target supplier evaluation system. This algorithm can help system users evaluate and classify suppliers more accurately, and provide a reliable basis for actual supplier management decisions.

Data preparation: Data related to supplier evaluation needs to be collected, while some supply chain-related information also needs to be collected.

Feature selection: According to the actual demand and available data, features related to supplier evaluation are selected, and these features can be continuous.

Data preprocessing: The collected data is preprocessed.

CART algorithm application: The CART algorithm is used to construct a supplier evaluation model. The CART algorithm is a decision tree algorithm that can classify and predict data based on its features and labels. By building an appropriate decision tree model, this paper can help to evaluate and select target suppliers.

Model evaluation and optimization: Cross-validation and other methods are used to evaluate the model, and check its performance and stability. If problems are found with the model, optimization can be achieved by adjusting the parameters of the decision tree or using other integrated algorithms.

4. Design of international trade target supplier evaluation system

4.1 System architecture design

The design of the evaluation system is one of the key factors to ensure that international trade target suppliers can better meet customer needs [18, 19]. The goal of system architecture design is to establish a stable, efficient, and scalable supplier evaluation system.

In the system architecture design, this paper introduces some advanced technologies and tools, such as cloud computing and big data analysis. These technologies help people process a large amount of data and provide real-time supplier evaluation results. The system architecture design is evaluated and optimized. Through continuous analysis of system performance and user feedback, the system is continuously improved to enhance evaluation effectiveness and user satisfaction [20].

4.2 Data processing and feature extraction

In the international trade target supplier evaluation system, data processing and feature extraction are one of the committed steps [21, 22]. This step aims to extract useful feature information from the original data for subsequent evaluation and analysis. When designing a data processing and feature extraction module, the following aspects need to be considered (as shown in Fig. 1).

Figure 1.

Data processing and feature extraction process.

Data cleansing and preprocessing: The original data is cleaned and preprocessed, including missing values, outliers, and duplicate values. Through these processes, the quality and accuracy of data can be improved, providing a reliable foundation for subsequent analysis.

Feature selection and extraction: In the process of data processing, appropriate features need to be selected according to the needs of the evaluation system. Through feature selection and extraction, the original data can be transformed into more representative and differentiated features, thus improving the effectiveness of the evaluation system.

Data conversion and standardization: For different types of features, appropriate data conversion and standardization are required. For example, for continuous features, normalization can be performed, and for classified features, encoding conversion can be performed. This can make different types of features comparable and better adapt to the modeling requirements of the evaluation system.

Feature engineering and optimization: During feature extraction, feature engineering and optimization can be carried out with the help of domain knowledge and professional skills. By combining, splitting, and deriving features, the expression ability and discrimination of features can be further improved, thereby improving the performance of the evaluation system.

In the design of an international trade target supplier evaluation system, data processing and feature extraction are important steps [23]. Through reasonable data cleansing, feature selection, and extraction, as well as data conversion and standardization and other operations, high-quality and differentiated features can be provided for subsequent evaluation and analysis, thus improving the accuracy and reliability of the evaluation system.

5. Implementation and evaluation of the evaluation system for international trade target suppliers

5.1 System implementation technology selection

System implementation and evaluation is an important part of the design of an international trade target supplier evaluation system, which involves selecting suitable technology implementation solutions and evaluating and verifying the system [24, 25]. In terms of system implementation technology selection, it is necessary to consider factors such as system reliability, performance, security, and scalability.

This paper can choose to use existing mature technology platforms for system development, such as cloud computing-based platforms or enterprise software development platforms. It is necessary to choose the corresponding development language and framework based on specific needs, such as Java, Python, etc. It is also necessary to consider the frontend and backend technologies of the system, including the selection of databases and the way of data storage and processing. In the process of system implementation, attention should be paid to the maintainability and modular design of the code to facilitate later modifications and maintenance. For the evaluation of the system, methods such as functional testing, performance testing, and security evaluation can be used to comprehensively inspect and verify the system, ensuring that it can meet the expected evaluation needs of trade target suppliers. At the same time, professional users or domain experts can also be invited for evaluation to obtain more comprehensive feedback and improvement suggestions. System implementation and evaluation is an iterative process that requires continuous optimization and improvement to ensure that the system can continuously and stably support the evaluation of international trade target suppliers [26].

5.2 System function implementation

The implementation of system functions aims to provide an easy-to-use and fully functional supplier evaluation system to meet the needs of users. To achieve this goal, this article has designed the following main functions:

Supplier information entry: Users can enter the basic information of suppliers into the system, including company name, contact person, address, and contact information. The system should have corresponding data verification and integrity check mechanisms to ensure the accuracy and completeness of the entered information.

Supplier evaluation indicator definition: The system should provide the function of defining evaluation indicators, and users can customize the evaluation indicator system based on actual needs. These indicators can include suppliers’ quality control capabilities, on-time delivery, and price competitiveness. Users can flexibly define evaluation criteria by adding, editing, and deleting evaluation indicators.

Supplier evaluation record management: The system should provide a management function for evaluation records, allowing users to create evaluation records in the system and evaluate different suppliers. Evaluation records can include information such as the score of evaluation indicators and evaluation time. Users can query and view past evaluation records at any time for comprehensive evaluation and comparison.

Analysis of evaluation results and report generation: The system should have the function of analyzing evaluation results and generating reports. By analyzing evaluation records, the system can generate a comprehensive report of supplier evaluations, helping users better understand supplier performance. The report can include the overall evaluation scores of suppliers, scores of various evaluation indicators, etc., and can be exported and printed as needed.

Through the implementation of the above functions, this article provides users with a comprehensive, flexible, and convenient supplier evaluation system to help people effectively manage and evaluate suppliers, and achieve the smooth implementation of international trade goals (as shown in Fig. 2).

Figure 2.

International trade target supplier evaluation system diagram.

5.3 System evaluation methods

During the evaluation process, it is necessary to determine the evaluation indicator system, including but not limited to supplier reputation, on-time delivery rate, product quality stability, etc. Supplier-related information needs to be collected, such as registration information, business status, quality certification, etc. The combination of quantitative analysis and qualitative analysis is used to evaluate suppliers, and information can be collected through methods such as data statistical analysis, questionnaire surveys, and supplier interviews. During the evaluation process, attention should also be paid to the authenticity and reliability of the data, and if necessary, third-party institutions can be used for data verification. Through comprehensive evaluation results, a comprehensive evaluation of suppliers is obtained to guide decision-makers in making accurate decisions when selecting suppliers. The application of system evaluation methods can provide a scientific basis for the improvement of the evaluation system for international trade target suppliers, and improve supply chain management and trade efficiency.

5.4 System test

Selecting a reliable supplier is essential to maintaining supply chain stability and business continuity in international trade collaboration. It is especially crucial to have a comprehensive target supplier evaluation system to assess suppliers’ overall performance. System testing has become crucial to ensuring the stability and accuracy of the evaluation system. The test material of the international trade target supplier assessment system will be presented in this paper.

Verifying if the evaluation system is operating in compliance with established functional requirements is the aim of system functional testing. The tester will mimic the actions of a user and carry out a thorough examination of all the functional modules of the system, including the creation and export of assessment results and the entry, inquiry, change, and deletion of supplier information, among other tasks.

The primary goal of performance testing is to assess the system’s ability to function at various loads. To make sure the system can continue to function steadily in high concurrency scenarios, a large number of users can be simulated accessing the system simultaneously, and its response time, throughput, number of concurrent users, and other performance indicators can be tested. Test of interface friendliness: The interface friendliness test primarily assesses how intuitive and simple the system’s user interface is to use. Verify that the system’s prompt information, operating procedure, interface design, and other elements align with user habits from the perspective of the user. Test for response time requirements: Following testing, it was discovered that the response time when a user logs in to execute any kind of operation is within 5 seconds, which satisfies requirements on both the client and management side. The purpose of compatibility testing is to confirm that the evaluation system works with various devices, browsers, and operating systems. To make sure the system can function normally in a variety of situations; testers can run it in diverse conditions. The process of evaluating an evaluation system’s resistance to possible security risks is called security testing. To guarantee the security and integrity of system data and to comply with requirements, the test content comprises system access control, data encryption, vulnerability scanning, and other measures. The availability test primarily assesses the assessment system’s learnability and usability to meet the requirements.

6. Performance testing of the evaluation system for international trade target suppliers

6.1 Dataset introduction

The selection of a dataset is based on collected international trade-related data, and evaluation and analysis are conducted on target suppliers. The data set includes the basic information of suppliers, product quality data, delivery time data, price data, etc. Through the introduction of datasets, people can understand key information such as the source, type, and scale of the data used. At the same time, data preprocessing methods, such as data cleansing and missing value processing, are also introduced to ensure the accuracy and integrity of data (as shown in Table 1).

Table 1
Introduction to dataset sources

Data name	Data sources	Data type	Data scale	Data preprocessing methods
Supplier basic information	Enterprise database	Text, numerical value	1000 records	Data cleansing and missing value processing
Product quality data	Quality inspection report and user feedback	Numeric value, text	5000 records	Data cleansing and missing value processing
Delivery time data	Order records, logistics tracking	Date, numerical value	2000 records	Data cleansing and Outlier handling
Price data	Procurement contract and invoice	Numerical value	3000 records	Data cleansing and Outlier handling

6.2 Experimental data

In the international trade target supplier evaluation system designed in this article, a series of experiments were conducted to verify the feasibility and effectiveness of the system.

(1) System evaluation level effect

This article collected a large amount of actual trade data and used the system to evaluate and analyze these data. Through the automation function of the system, various indicator data of suppliers can be quickly and accurately obtained, and they can be comprehensively evaluated and ranked. This article selected ten suppliers and evaluated their quality indicators, delivery time indicators, price indicators, and comprehensive evaluation levels. The specific analysis results are shown in Fig. 3. Among them, the comprehensive evaluation level was measured by the average score of the three indicators, as shown in Table 2.

Table 2
Supplier comprehensive evaluation levels

Score interval	Grade	Supplier
[0, 60)	Terrible	/
[60, 70]	Poor	5, 6
(70, 80)	Commonly	3, 9
[80, 90)	Good	1, 4, 7, 10
[90, 100)	Excellent	2, 8

Figure 3.

Supplier-level evaluation effect.

As shown in Fig. 3 and Table 2, this article took ten suppliers as the research objects and researched the scores of quality indicators, delivery time indicators, price indicators, and comprehensive evaluation levels. These data were based on actual trade data and evaluated using the international trade target supplier evaluation system.

Comprehensive evaluation level: The comprehensive evaluation level is the result of a comprehensive calculation based on the scores of various indicators. Supplier 2 and Supplier 8 had the highest comprehensive evaluation, both receiving excellent ratings. The comprehensive evaluation level of Supplier 1, Supplier 4, Supplier 7, and Supplier 10 was good. The comprehensive evaluation level of Supplier 3 and Supplier 9 was average. The comprehensive evaluation level of Supplier 5 and Supplier 6 was relatively low.

Through the analysis of the experimental results, it can be seen that the performance of suppliers on different indicators is good or bad, as well as their comprehensive evaluation level. The comprehensive evaluation level can help people evaluate the capabilities and performance of suppliers more comprehensively. Based on these evaluation results, it is possible to select and collaborate with the best suppliers, optimize supply chain management, and provide a scientific basis for subsequent international trade decisions.

(2) Model comparison

The dataset was input into the traditional algorithm and the CART algorithm respectively, and the accuracy, precision, and recall performance of the two algorithms were compared. The traditional algorithm used in this article is the Naive Bayes algorithm (A algorithm), and the CART algorithm is the Decision Tree algorithm (B algorithm). In model comparison, it is necessary to adjust algorithm parameters to improve model prediction accuracy. The experimental results are shown in Fig. 4.

Figure 4.

Comparison of model accuracy under two different algorithms.

As shown in Fig. 4, Fig. 4a shows the model accuracy under the traditional algorithm, and Fig. 4b shows the model accuracy under the CART algorithm. This article compared the accuracy, precision, and recall of traditional algorithms and CART algorithms, and the results were as follows.

Accuracy: The ability of the model to correctly classify samples was measured. In traditional algorithms, the accuracy range of suppliers ranged from 0.77 to 0.92, with an average accuracy of 0.83. In the CART algorithm, the accuracy range of suppliers ranged from 0.83 to 0.94, with an average accuracy of 0.88. It can be seen that the accuracy of the CART algorithm was slightly higher than that of traditional algorithms.

Precision: The proportion of samples predicted by the model to be positive was measured as the true positive case. In traditional algorithms, the accuracy of suppliers ranged from 0.79 to 0.93, with an average accuracy of 0.85. In the CART algorithm, the accuracy of suppliers ranged from 0.83 to 0.95, with an average accuracy of 0.89. It can be seen that the accuracy of the CART algorithm was slightly higher than that of traditional algorithms.

Recall rate: The proportion of samples correctly predicted as positive cases by the model to the total true positive cases was measured. In traditional algorithms, the supplier’s recall rate ranged from 0.76 to 0.91, with an average recall rate of 0.82. In the CART algorithm, the supplier’s recall rate ranged from 0.81 to 0.93, with an average recall rate of 0.87. It can be seen that the recall rate of the CART algorithm was slightly higher than that of traditional algorithms.

For the given dataset, the CART algorithm performed better than traditional algorithms in terms of accuracy, precision, and recall.

(3) Model tuning

The best algorithm was selected for subsequent model parameter adjustment. The maximum depth and the minimum sample number of leaf nodes are two important parameters in the decision tree model, which have a great impact on the accuracy and complexity of the model. Therefore, it is necessary to adjust these two parameters and record the model training time to find the optimal model. This article tested the training time of traditional algorithms and CART algorithms. The experimental results are shown in Fig. 5.

Figure 5.

Training time under different algorithms after model parameter tuning.

Figure 6.

System application evaluation scores under two algorithms.

As shown in Fig. 5, it can be observed that the training time of the traditional algorithm and the CART algorithm was roughly similar. The training time range of the traditional algorithm was from 0.4 seconds to 1.3 seconds, with an average training time of 0.85 seconds; The training time range of the CART algorithm was from 0.5 seconds to 1.5 seconds, with an average training time of 0.98 seconds. It is difficult for this article to accurately predict the impact of maximum depth and minimum sample size of leaf nodes on training time like traditional algorithms. Therefore, when further adjusting parameters, methods such as cross-validation can be used to select the optimal parameter combination to achieve high model accuracy and appropriate model complexity.

(4) System application

In the system application, this article compared the application effectiveness rating, supplier ranking level, and system user satisfaction rating of traditional algorithms and CART algorithms. In real business scenarios, these indicators are crucial for determining whether the model is suitable for practical applications. Based on the results of this experiment, the optimal algorithm and parameter combination can be selected for subsequent system development and application. The test results are shown in Fig. 6.

As shown in Fig. 6, Fig. 6a shows the system application evaluation score under the CART algorithm, and Fig. 6b shows the system application evaluation score under the traditional algorithm. The specific situation is as follows:

For traditional algorithms, the application effectiveness rating range was between 3.8 and 4.9, with an average rating of 4.3. Among them, Supplier 4 received the highest score of 4.9, while Supplier 3 and Supplier 10 received the lowest score of 3.8. For the CART algorithm, the application effectiveness rating range was between 4.2 and 5.0, with an average rating of 4.6. Among them, Supplier 4 received the highest score of 5.0, while Supplier 3 and Supplier 7 received the lowest score of 4.2. From the perspective of application performance evaluation, the CART algorithm performed better overall, with higher scores and a smaller scoring range.

In traditional algorithms, the user satisfaction rating range of the system was between 3.5 and 4.9, with an average rating of 4.2. In the CART algorithm, the system user satisfaction rating range was between 4.0 and 5.0, with an average rating of 4.5. From the perspective of system user satisfaction ratings, the CART algorithm performed better overall, with higher ratings and a smaller rating range.

Based on these data, this article can conclude that the CART algorithm performs better in indicators such as application effectiveness rating, supplier ranking level, and system user satisfaction rating, with higher ratings and more stable results. Therefore, the CART algorithm can be chosen as the optimal algorithm and appropriate parameter combinations can be selected based on specific needs.

7. System optimization and application prospects

7.1 System performance optimization

Optimization of system response time: By evaluating and analyzing the response time of various modules in the system, performance bottlenecks are identified and optimized to improve the system’s response speed and user experience.

Database performance optimization: The database in the system is subject to performance optimization, including index design and optimization, query statement optimization, database cache optimization, etc., to improve the system’s ability to process big data.

Optimization of network transmission performance: Various aspects of network transmission in the system are optimized, including network bandwidth optimization, data compression, and encryption optimization, to improve the performance of the system in the network environment.

Optimization of concurrent access: In response to the concurrent access pressure that the system may face, optimization of concurrent access is carried out, including the design and optimization of distributed architecture, optimization of load balancing, optimization of caching mechanism, etc., to improve the system’s concurrent processing ability.

Optimization and utilization of system resources: Various resources in the system are optimized and utilized, including memory management and optimization, CPU (Central Processing Unit) utilization optimization, disk space optimization, etc., to improve the overall performance and resource utilization of the system.

Through research and application of system performance optimization, the design of the “International Trade Target Supplier Evaluation System” can be more efficient and stable in practical applications to meet user needs and provide a foundation for further application prospects.

7.2 System application prospects

In terms of optimizing the evaluation system for international trade target suppliers and its application prospects, the following key points can be considered.

The optimization of the system can be carried out from the perspective of data collection and processing, by introducing more advanced technologies and algorithms to improve the accuracy and reliability of data. It can further improve the functions and modules of the system, add more evaluation indicators and weight allocation methods, and make the system more comprehensive and detailed. It can also be considered to introduce machine learning and artificial intelligence technologies to achieve automated evaluation and prediction functions and improve the intelligence level of the system. In terms of application prospects, the system can be widely applied in various fields of international trade, providing more accurate and reliable evaluation results for enterprises and suppliers, and helping enterprises choose suitable suppliers and optimize supply chain management. At the same time, the application of the system can also promote the standardization and transparency of international trade and improve the efficiency and competitiveness of the entire trading system. Therefore, through system optimization and further application prospects, it can be able to play a greater role in the field of international trade, providing better services and support for enterprises and suppliers [27].

7.3 Continuous improvement and development

In the design of an international trade target supplier evaluation system, system optimization, and application prospects are important aspects. Among them, continuous improvement and development are a key element in the optimization and application prospects of the system. Through continuous improvement and development, the performance and functionality of the system can be continuously improved to meet the constantly changing international trade needs.

Continuous improvement refers to the continuous improvement and optimization of a system during operation based on user feedback and market demand. This can include using advanced technological means to improve system stability and security, increase system adaptability and flexibility, and better respond to various complex international trade scenarios.

Continuous development refers to the continuous introduction of new functions and services based on improved system functionality to meet the growing needs of users. This can include increasing integration with other related systems, providing more data analysis and decision support functions, and improving user interfaces and experiences to enable users to use the system more conveniently and quickly.

Through continuous improvement and development, the international trade target supplier evaluation system can continuously adapt to and meet the constantly changing international trade environment and needs. Meanwhile, continuous improvement and development can also enhance the competitiveness of the system, and attract more users to use the system, further promoting the development of international trade and improving the efficiency of supply chain management.

8. Conclusions

Through performance testing and data analysis of the evaluation system for international trade target suppliers, the following conclusions can be drawn. When evaluating and ranking suppliers, the system can quickly and accurately obtain various indicator data of suppliers, and conduct comprehensive evaluation and ranking, thereby helping to select and cooperate with the best suppliers. In model comparison, the CART algorithm performs better in terms of accuracy, precision, and recall compared to traditional algorithms, with higher ratings and more stable results. In model parameter tuning, the training time of traditional algorithms and CART algorithms is similar, and further methods such as cross-validation need to be used to select the optimal parameter combination. According to the system application evaluation score, the CART algorithm performs better in indicators such as application effectiveness score, supplier ranking level, and system user satisfaction score, with higher scores and a smaller scoring range. The international trade target supplier evaluation system has demonstrated good performance and effectiveness in experiments, which can help select excellent suppliers, optimize supply chain management, and provide a scientific basis for subsequent international trade decisions in practical applications.

The design and research of the international trade target supplier evaluation system based on the CART algorithm faces some challenges and future development trends. Since the accuracy and effectiveness of the evaluation system are highly dependent on the quality of the data, supplier data may be incomplete, inaccurate, or outdated. At the same time, supplier evaluation involves multiple dimensions and indicators, such as quality, delivery punctuality, cost, etc., How to comprehensively consider these indicators and make trade-offs is a challenge. For the existing limitations and problems, multiple algorithms can be integrated in the future, and the CART algorithm can be combined with other machine learning algorithms such as random forests and neural networks to improve the accuracy and generalization of the evaluation system. With the development of big data technology, big data analysis technology can be used to process supplier data and mine potential information and insights. In summary, the international trade target supplier evaluation system based on the CART algorithm faces challenges and also has broad prospects for development. It needs to continuously optimize the algorithm and introduce new technologies to improve the performance and effectiveness of the system.

References

Wiȩckowski

Kizielewicz

Shekhovtsov

Sałabun

. How do the criteria affect sustainable supplier evaluation? – A case study using multi-criteria decision analysis methods in a fuzzy environment. Journal of Engineering Management and Systems Engineering. 2023; 2(1): 37-52.

Hazza

Abdelwahed

Ali

Sidek

ABA

. An integrated approach for supplier evaluation and selection using the Delphi method and analytic hierarchy process (AHP): A new framework. Int. J. Technol. 2022; 13(1): 16-25.

Yazdani

Torkayesh

Chatterjee

. An integrated decision-making model for supplier evaluation in the public healthcare system: the case study of a Spanish hospital. Journal of Enterprise Information Management. 2020; 33(5): 965-989.

Liou

JJH

Chang

Hsu

. Application of an MCDM model with data mining techniques for green supplier evaluation and selection. Applied Soft Computing. 2021; 10(9): 107534.

Bai

Kusi-Sarpong

Badri Ahmadi

Sarkis

. Social sustainable supplier evaluation and selection: A group decision-support approach. International Journal of Production Research. 2019; 57(22): 7046-7067.

Chatterjee

Stevic

. A two-phase fuzzy AHP-fuzzy TOPSIS model for supplier evaluation in a manufacturing environment. Operational Research in Engineering Sciences: Theory and Applications. 2019; 2(1): 72-90.

Amiri

Hashemi-Tabatabaei

Ghahremanloo

Keshavarz-Ghorabaee

Zavadskas

Banaitis

. A new fuzzy BWM approach for evaluating and selecting a sustainable supplier in supply chain management. International Journal of Sustainable Development & World Ecology. 2021; 28(2): 125-142.

Bagante

Spolverato

Merath

Weiss

Alexandrescu

Marques

. Intrahepatic cholangiocarcinoma tumor burden: A classification and regression tree model to define prognostic groups after resection. Surgery. 2019; 166(6): 983-990.

Kori

Mahabaleshwar

. Classification and regression tree (CART) based resource allocation scheme for wireless sensor networks. Computer Communications. 2023; 19(7): 242-254.

10.

Carrizosa

Molero-Río

Morales

. Mathematical optimization in classification and regression trees. Top. 2021; 29(1): 5-33.

11.

Gomes

ACM

Lemos

Jelihovschi

. Comparing the predictive power of the CART and CTREE algorithms. Avaliação Psicológica. 2020; 19(1): 87-96.

12.

Dusseldorp

Meulman

. Multiple moderator meta-analysis using the R-package Meta-CART. Behavior Research Methods. 2020; 5(2): 2657-2673.

13.

Cheng

Zhu

. Research on performance evaluation of planning and design projects based on dea-cart algorithm – taking a research institute’s planning and design projects as an example. Financial Communication: Comprehensive Edition. 2021; 000(022): 94-98.

14.

Cheng

Song

Chen

Cheng

. Intelligent safe driving methods based on hybrid automata and ensemble CART algorithms for multi high-speed trains. IEEE Transactions on Cybernetics. 2019; 49(10): 3816-3826.

15.

Bian

. Research on intelligent operation system for supplier evaluation driven by big data. Modern Management. 2020; 10(2): 298-303.

16.

Sun

Chen

. Zinc ore supplier evaluation and recommendation method based on nonlinear adaptive online transfer learning. Journal of Industrial and Management Optimization. 2023; 19(1): 472-490.

17.

Yurtay

Ayanolu

. Dashboard application model in supplier evaluation by using artificial immune system and data mining methods. Pamukkale University Journal of Engineering Sciences. 2021; 27(2): 162-172.

18.

Cheng

Peng

Zhang

Liu

Zhou

Yang

Huang

. An intelligent supplier evaluation model based on data-driven support vector regression in the global supply chain. Computers & Industrial Engineering. 2020; 139(Jan): 105834.1-105834.11.

19.

Mohammad

Farjood

. Supplier evaluation/selection in behavioral aspects consideration. Siddhant – A Journal of Decision Making. 2020; 20(3&4): 73-86.

20.

Mowri

Siddula

Roy

. Is iterative feature selection technique efficient enough? A comparative performance analysis of RFECV feature selection technique in ransomware classification using SHAP. Discov Internet Things. 2023; 3(21).

21.

Guo

. Evaluation index extraction and system construction of grid material supplier: A meta-analytic review. Smart Grid. 2018; 08(1): 39-60.

22.

Bana

Michnik

. Evaluation of the impact of strategic offers on the financial and strategic health of the company – a soft system dynamics approach. Mathematics. 2019; 7(2): 208-208.

23.

Ahmed

Kusi-Sarpong

Arhin

Kusi-Sarpong

. Supplier sustainability performance evaluation and selection: A framework and methodology. Journal of Cleaner Production. 2018; 205(PT.1-1162): 964-979.

24.

Suvrat

Kim

Linderman

. An empirical investigation of transaction dynamics in online surplus networks: A complex adaptive system perspective. Journal of Operations Management. 2019; 65(2): 160-189.

25.

Leyla

Enes

Ceyhun

Ali

. A green supplier evaluation system based on a new multi-criteria sorting method: VIKORSORT. Expert Systems with Application. 2018; 114(DEC): 479-487.

26.

Coquidé

Lages

Dima

. Prospects of BRICS currency dominance in international trade. Appl. Netw. Sci. 2023; 8(1): 65.

27.

Nagurney

. Food security and multicommodity agricultural international trade: Quantifying optimal consumer subsidies for nutritional needs. Int. Trans. Oper. Res. 2024; 31(3): 1375-1396.

Design of an international trade target supplier evaluation system based on CART algorithm

Abstract

Keywords

1. Introduction

2.1 Supplier evaluation system

2.2 CART algorithm

3. Data processing based on CART algorithm

4.1 System architecture design

4.2 Data processing and feature extraction

5.1 System implementation technology selection

5.2 System function implementation

5.4 System test

6. Performance testing of the evaluation system for international trade target suppliers

6.1 Dataset introduction

Table 1 Introduction to dataset sources

(1) System evaluation level effect

Table 2 Supplier comprehensive evaluation levels

(2) Model comparison

(3) Model tuning

(4) System application

7.1 System performance optimization

7.2 System application prospects

7.3 Continuous improvement and development

8. Conclusions

References

Table 1
Introduction to dataset sources

Table 2
Supplier comprehensive evaluation levels