Naive Bayes Phishing Detection Based On Feature Selection And Reinforcement Training

With the rapid development of communication technology and the popularization of networks,the amount of data that people generate daily is increasing.The demand for how to extract useful information from data is growing.In recent years,data mining technology has received widespread attention from researchers.Data mining is a technique that try to discover the connection among a large amount of data,and find out the information hidden behind the data.Classification algorithm is a branch of data mining.The classification algorithm is mainly to design the classifier through the algorithm,and use classifier to divide the data items in the data set into a fixed classification.Bayesian algorithm is a classic classification algorithm based on probability statistics.Because of its simple method,high classification accuracy,and fast speed,it is widely studied and used.Bayesian algorithms mainly include Naive Bayes algorithm and Tree Augmented Naive Bayes algorithm.This paper mainly researches on Naive Bayes algorithm.(1)Using the concept of information entropy,this paper proposes a data preprocessing method that uses information gain for feature selection.Through this method,more valuable features can be selected to build a classification model,which is convenient for adjusting and improving the accuracy of the classifier.Because fewer features are selected,the training efficiency of the classification model will be improved.It takes less time to train the model.(2)This paper proposes a Reinforcement training method for training methods.Divide the training set of the data set,After training by use one piece of data,then predict the next piece of data.Adjust the training model by the wrong prediction.In this way,the classifier can be trained with fewer data items and faster speed.On this basis,this paper proposes a phishing website detection system FSRT-NB(Feature Selection and Reinforcement Training-Naive Bayesian)based on a naive Bayes classifier for feature selection and Reinforcement training.Experimental results show that FSRT-NB classifier has good classification effect and efficiency.