目的機能から基本機能へ (5)

―予測の品質工学, TS法―

抄録

This chapter deals with various kind of prediction problems. Traditional ways of prediction are either physical models or the use of statistical regression. There has been no good mathematical model for earthquake prediction. The statistical method using multivariate analysis is a least square method. It is good only fbr the data used in the least square. Dr. C. R. Rao, a famous multivariate specialist stated in his book that any conclusion used is his book must be checked with different data. This chapter explains the method called Taguchi-Schmidt methodology for various kind of prediction problems. It is the methods to use the Schmidt expansion among all variables whithin unit group. For an example in case of prediction of total sales of a company, choose all companies whose total sales of this year was almost same as those of last year. It is called unit group. Various data of all companies should collected. After Schmidt expansion among all variable within unit group calculate the difference of variable of all companies. Then calculate error of predition of each variable using linear slope explained in this article.