Rough Fuzzy Regression Algorithm And Its Application

Regression analysis has become a social science quantitative ana-lysis of a data processing method, which is the most basic, the most widely used. It can not only describe the function expression between the independent variables and the dependent variable correlation, but also be used to predict the dependent variable. However, according to the values of the independent variable prediction of a sample of the results can only have a value. According to current scientific theory, the predicted value is often uncertain, but it will change in a range. Therefore, we introduce a new range algorithm to give the prediction values, that is to say, the value prediction of upper and lower approximation.In this paper, we will put the traditional regression analysis and fuzzy rough theory together, describe a regression algorithm based on Fuzzy Rough Set, including the use of neighborhood rough set attribute reduction, for fuzzy similar matrix, fuzzy approximation and regression prediction. Different from the traditional regression, using this algorithm eventually get two predictive regression curve, and attribute values fall right into between the two prediction curve. UCI data experiments are carried out to validate the above algorithm.The content of this thesis includes the following two parts:In the first, it constructs attribute reduction based on neighborhood problems according to the attribute reduction based on the classic theory of rough set. It presented the dependence and significance of attribute concept under the neighborhood rough set decision system, and tested with actual data.Secondly, the article analyzed the traditional regression analysis and main steps of characteristics, combined regression analysis theory with rough set theory, proposed a fuzzy rough regression algorithm. It com-pared with the traditional regression analysis of difference, and put forward the fuzzy rough regression algorithm. Last the algorithm is verified according to the actual data.