Algorithm Research On Spam Instant Message Detection In Mobile Internet

With the rapid development of mobile Internet and the popularization of mobile intelligent terminals,people often suffer a lot of rubbish information while experiencing the convenience of mobile communication.At present,the garbage information spread in various media forms with the aid of convenient instant communication,which cause great harm to society.In order to create a good communication environment,the research of accurate and efficient spam detection and filtering method is always the focus of people's attention.Aiming at the characteristics of spam instant message in mobile Internet environment,this paper designs and realizes the spam instant message detection based on naive Bayesian method.Firstly,the collected garbage instant message data set is divided into training set and test set,respectively preprocessing,and the instant message text feature of vector is used to train naive Bayesian classifier,and then the test set is entered into the training classifier to classify.The experimental results show that the method of garbage instant message detection based on naive Bayesian algorithm has short training time and simple implementation.Because the method of spam instant message detection based on naive Bayesian algorithm ignores the connection between instant message text features,which leads to the low classification accuracy and recall rate.Making the use of the ability that deep belief network model could fully excavate the correlation between text features,the paper designed and achieved a text classifier based on deep belief network model to realize the detection of spam instant message.The process is firstly preprocessing the spam instant message,then using the information gain(IG)method to extract the effective features,comparing the importance of the selected features with TF-IDF weight calculation,and representing the character of the text by using the Boolean logic model.Finally,putting the obtained eigenvector to the input layer of the deep belief network to train the classifier,in which the number of nodes in the output layer is determined by the category label number,and then using the trained classification model to classify the test samples.The experimental results show that,compared with naive Bayesian algorithm,the classification model based on deep belief network can excavate the deeper information of instant message,and the ability of distinguishing the spam instant message is stronger,the recognition accuracy is higher,and it can bemore adapt to the changing text form of spam instant message.