Automatic Classification Of Portal News Based On Machine Learning

With the surging of portal news and the increasing complexity of news content,traditional manual methods of news classification have become less and less efficient.Meanwhile,manual classification of news is prone to be influenced by subjective factors,thereby decreasing classification accuracy.Therefore,traditional manual methods of news classification have failed to meet the demands of the new age.To this end,there is a need to study an automatic classification method suitable for portal news to solve the existing problems.The machine learning-based approach emerges as the times require,which is expected to enable automatic classification of portal news and solve the shortcomings of manual news classification.This thesis deeply analyzes the text characteristics of portal news,and on the basis of studying the key technologies of machine learning applied in text classification,compares several commonly-used techniques in the text classification of news through experiments.Ultimately,a machine learning-based method that is suitable for the automatic classification of portal news is proposed,and its design idea and overall design scheme is put forward.This thesis first pre-processes news texts,conducts feature extraction with the CHI method,and then performs weight calculation of the features with TF-IDF,and finally builds a classification model with the SVM classification algorithm.Ultimately:,this thesis develops a machine learning-based automatic classification system of portal news for the Exploration and Production Branch of China Petroleum.Through the system,the news obtained from news sources can be automatically classified into the category to which it belongs.Experimental results show that the system proposed secures a classification accuracy rate of 85.1%,realizing the automatic classification of portal news,improving classification efficiency,reducing enterprise cost,and meanwhile greatly uplifting classification accuracy.