Interval-valued test cost sensitive attribute reduction related to risk attitude

• Published in: International journal of machine learning and cybernetics Vol. 15; no. 9; pp. 4155 - 4174
• Main Authors: Lu, Yaqian, Liao, Shujiao, Yang, Wenyuan, Guan, Ya’nan, Wu, Di
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2024; Springer Nature B.V
• Subjects: Algorithms; Approximation; Artificial Intelligence; Attitudes; Biochemistry; Biopsy; Breast cancer; Complex Systems; Computational Intelligence; Control; Cost analysis; Costs; Decision making; Engineering; Entropy; Genetic algorithms; Heuristic; Heuristic methods; Mammography; Mechatronics; Medical screening; Optimization; Optimization algorithms; Original Article; Pattern Recognition; Performance evaluation; Robotics; Rough set models; Set theory; State-of-the-art reviews; Systems Biology; Ultrasonic imaging; Uncertainty analysis; Intervals; Test cost; Uncertainty; Attribute reduction; Rough set
• ISSN: 1868-8071; 1868-808X
• DOI: 10.1007/s13042-024-02140-4

Attribute reduction is a typical topic in the field of rough sets. As an extension of this topic, test cost sensitive attribute reduction has garnered considerable attention in recent years, and scholars have made many achievements. However, existing research commonly operates under the assumption that test costs are exact values, disregarding the challenges associated with quantifying test costs accurately in certain real-world contexts. In light of the situation, this paper employs the form of interval values to represent the possible range of test costs and then studies the problem of attribute reduction based on interval-valued test costs. Firstly, a theoretical model for interval-valued test cost sensitive attribute reduction is constructed by utilizing a ranking method of intervals. In this model, some important concepts and properties are discussed. Especially, considering that the risk attitudes of different decision-makers may affect the decision results, an optimization problem related to risk attitude is formulated. Secondly, a backtracking algorithm and a heuristic algorithm are developed for tackling the optimization problem, along with the application of a competition strategy to enhance the heuristic algorithm’s performance. Finally, the performance of the two algorithms is evaluated on multiple UCI datasets, and comparisons are drawn with several state-of-the-art attribute reduction methods. Experimental analyses well illustrate the effectiveness and superiority of the suggested algorithms. It is hoped that this work provides new insights into cost sensitive attribute reduction from the perspective of cost uncertainty and provides a reference for decision-making problems.