Improvement Research Based On Data Mining Classification Algorithm

Driven by the trend of economic globalization,information technology has been developing rapidly,and a lot of data have been pouring out of every corner of our life.By manual analysis of human beings,the rich information which hidden behind these data can not be detected.In order to dig information from such a large amount of data,people urgently need to find a reasonable solution to deal with the challenges of excessive information.It is an important problem to study how to find useful information from a mass of data.The data mining technology transforms the original data into a suitable form of operation,extracts useful data for mining,and finally uses various strategies of data mining to generate useful patterns and rules.Among the methods of data mining,classification technology occupies a very important position.There are many popular classification algorithms,including decision tree classification algorithm,naive Bayes classification algorithm,neural network algorithm,SVM algorithm and K-nearest neighbor algorithm and so on.This thesis studies the classification algorithm,and proposes two improvement strategies and related implementation methods.(1)Before the training of data is started,the preprocessing part of the original data is improved,that is to standardize the magnitude of the original data,so that the values of the data are in the same order of magnitude,and then the obtained data are operated.This method can make data preprocessing better,and directly affect the effect and accuracy of subsequent data mining process,and give relevant experimental validation results.(2)Taking the least mean square algorithm and the BP neural network classification algorithm as the basic algorithms,using the rules of batch learning and adding the momentum factor,an improved batch learning BP algorithm is designed.The related experimental data shows that the improved batch learning BP algorithm can solve the classification problem better.