A method-numeric genetic algorithm (NGA) combined with nonlinear PLS (NPLS) is proposed in this paper, NPLS is improved and the obvious prediction model can be obtained, This algorithm can deal with all kinds of nonlinear function relations, Exponential, logarithmic, reciprocal and sigmoid functional formula are obtained, It is especially suitable to deal with the complicated nonlinear problem - complex structure and property relations, In this paper it is applied to solve the fungicidal activity of aseries of O-ethyl-N-isopropyl phosphoro(thioureido)thioates. It is compared with stepwise regression and PLS. The results of the correlation coefficient (R) and the standard residual (S-y) are all improved, A stable prediction can be obtained. NPLSNGA is proved to be promising in multivariate statistical method for quantitative structure-activity relationship analysis.
