To the characters of the data gathered from chemical industry, this study is on data mining technology applied in optimization of chemical products formula. Through study, we get the most suitable method to solve the problems in data processing, such as filling empty items in data set, checking and eliminating exceptional data, analyzing the principal component. Using the data from simulating a complicated nonlinear function, we compare several kinds of improved algorithms of BP neural network and analyze the main factors that influence the precision of neural network predicting. On the basis of above-mentioned work, we improve the traditional GA-ANN algorithm and realize the automatic search for BP neural network structure. Finally, we give some suggestion on using the technology to the optimization of chemical products formula in chemical industry.
|