Research On Data Decision-making Method Based On Rough Effect

Variable precision rough set is a common tool for data processing and decision making,and how to obtain the upper(lower)approximation set with the best approximation effect has important theoretical and application values for designing data processing methods and attribute reduction methods.In this paper,we take the different values of " misfound" and " not found" biases that must be faced in the process of data decision making as the background,based on the variable precision rough set model,and focus on the best approximation of the target set,we mainly do the following work: 1)A utility bias metric model(abbreviated as w-UB)that accommodates the value of bias is constructed,the basic properties of w-UB and the optimal approximation problem with a single set as the target set are discussed,the specific form of the optimal upper(lower)approximation set based on w-UB and the range of values of the best rough precision are given,and the characteristics of the best upper(lower)approximation set are further analyzed with specific arithmetic examples;2)The best approximation problem for multiple objective sets in the w-UB sense is discussed,and the best approximation metric model for multiple objective sets in the w-UB sense is given;3)A best approximation based attribute reduction method(abbreviated as δ(?)w-BAR)is proposed for decision information systems,with the best upper(lower)approximation set of the decision class as the core concern indicator;4)The connection and difference between δ(?)w-BAR and several common reduction methods are discussed,and the characteristics and performance of δ(?)w-BAR are analyzed from different perspectives with specific arithmetic examples and eight commonly used UCI datasets.The theoretical derivation and the analysis of arithmetic cases show that w-UB has good interpretability and the discussion of the best upper(lower)approximation set in relation to the rough precision can provide theoretical support for the construction of data processing methods with different bias values.Simulation experiments show that δ(?)w-BAR is easier to obtain a reduction set with fewer attributes than traditional algorithms such as divisional reduction and maximum distribution reduction,and can be used to integrate decision awareness into the data decision process in a simple way.Therefore,the discussion in this paper is a supplement and improvement of the existing decision making methods and attribute reduction methods,and has wide application value.