Fuzzy Linear Regression Prediction
Wu Chong, Pan Qishu, Li Hanling
(Harbin Institute of Technology,Harbin 150001,China)

Abstract: The dynamic prediction problem with fuzzy information is solved by using the linear regression prediction method. This involves the determination of fuzzy coefficients which are symmetric triangular fuzzy numbers and the definition of fitness. Expressions are given to describe the closeness of the estimated value and observed value of the equation for two symmetric triangular fuzzy numbers. The solution is expressed in term of a pair of functions expressed in cutset for symmetric triangular fuzzy numbers. More specifically, fuzzy coefficients are minimized under the restrictive condition that the fitness of the estimated value of each equation is not less than the predesignated fitness, while determination of fuzzy coefficients is reduced to solving a linear planning problem.Examples show that the model can yield results with very high accuracy.
Keywords: symmetric triangular fuzzy numbers;fuzzy linear regression;fuzzy prediction