Research And Realization Of Digitial Information Processing Technology Based On Machine Learning

The digital signal processing (DSP) already obtained wide application in many domains in the past several dozens years. But the DSP technology also has very many limitations along with the computer and the information technology development. For instance the DSP technology lacks the ability to do what one would like when facing the rich kinds of digital information. Therefore people need to extend and expand DSP technology in order to satisfy the new request. It has the important research value that people applies the machine learning as one kind of mature DSP technology in digital information processing.The paper take email the one kind of common digital information mediums as the research object. It realizes the intelligent processing of the mail digital information through the research of mail digital information processing based on machine learning. It enhances the classification and recognition accuracy of the mail digital information in order to satisfy the request that the mail gateway accurately filters the spam mail.This paper analyzed the digital information processing technology and the application of machine learning in digital information processing firstly. And it studied how to apply the machine learning technology in mail digital information processing. Then it studied and designed the feature selection algorithm and the classify algorithm of mail digital information. Mail digital information feature space dimension is very huge. In order to reduce the feature dimension and caused the machine learning algorithm to be feasible, it has utilized feature selection algorithm based on the feature selection measure function. And on the foundation it has researched and designed the feature selection algorithm which combines the genetic algorithms and the feature selection measure function. At the same time, it has designed the BP nerve network classifier in order to realize intelligent processing of mail digital information and enhances the classified recognition effect. And it optimized the BP neural network by genetic algorithms and has realized the GABP network classifier. Finally, the paper designed and realized an email information...