Feature selection and information fusion based on preference ranking organization method in interval-valued multi-source decision-making information systems

Abstract

Multi-source Decision-Making Information Systems (DMSs) demonstrate superior capabilities in integrating and analyzing a diverse array of information sources, providing enhanced functionality over single-source systems. Within these systems, feature selection is crucial for identifying key attributes, which reduces information and enhance the efficiency of the decision-making process. However, current established information fusion techniques in multi-source DMSs, which integrate various sources into a unified framework, tend to be computationally intensive and are not adept at handling interval-valued data. This paper introduces an innovative feature selection model specifically developed for multi-source DMSs, employing the Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE). The model initiates by establishing the neighbourhood relationships among objects across different attributes. It then utilizes the PROMETHEE algorithm to rank these attributes based on their comparative strengths and weaknesses, facilitating the pinpointing of the most valuable features. The model further refines the selection process by quantifying the consensus level, thereby discovering the most reliable information sources. Our some experiments, performed utilizing a broad and comprehensive dataset, have validated both the model and its underlying algorithm. The results obtained provide compelling evidence of the model's effectiveness, especially highlighting its proficiency in handling interval-valued data. Furthermore, the outcomes illustrate the model's significance to the enhancement of decision-making processes within multi-source Decision-Making Information Systems (DMSs).