E-mail classification algorithm based on user s action
Chen Zhi-pin · Journal of Computer Applications · 2014
It is difficult to build a personalized classification model to filter spam,and the model is also difficult to adapt to the user changing interest.To solve these problems,a novel E-mail classification algorithm based on users' action was proposed.With analyses of Naive Bayesian(NB) algorithm,the computing procedure was re-built to represent the dynamic adjustment abilities of the classification.It automatically classified the new E-mails received in the mail server.With the user's action on the E-mails,the system collected the mis-classified E-mails into the training set,and updated the features' frequencies.Therefore,the model for E-mail classification was automatically adjusted to filter the junk E-mails more effectively.Using a small set of sample data as training set and comparing with Naive Bayesian and risk-sensitive NB,the experimental results show that the recognition rate of the new algorithm has improved over 10%.