Research On Text Analysis Of Current Political News Based On Machine Learning

With the maturity of the "Internet+" model and the rapid development of information technology in recent years,the Party Central Committee and local governments at all levels have incorporated the process of using information technology for spiritual civilization construction into the work requirements,the main program of spiritual civilization construction.It is to carry out social management innovation work,which is embodied in the publicity work of government affairs.How to scientifically and rationally use information technology to build spiritual civilization has become a hot issue about strengthening social management innovation and improving the content of government affairs publicity.Therefore,under the influence of the time span brought by the staged update of government affairs propaganda work,the analysis of the hot spots of current affairs can still maintain its scientific and time-sensitive characteristics and become the key point to solve the problem.this article will analyze the news texts of government affairs propaganda work through data mining,text classification,neural network and other algorithms from the perspective of combining machine learning technology,providing scientific and effective analysis methods for social management innovation.This article uses the news text on the "learning power" news platform as a data source,and analyzes 5058 news texts from July 2018 to December 2019 for classification and processing.Taking the 10 known classification categories as the classification target the article uses text processing technology,the construction of the classification model,and the neural network algorithm to complete the optimization of the hot analysis of current affairs news,and visually compares and shows the difference before and after optimization.Hotspots are automatically updated periodically.Based on this background,this article has completed the following aspects of work:(1)Use the web crawler technology to crawl the news text of the "Learning Power"website.In this way,a news text corpus is constructed to lay a foundation for the classification of news texts.(2)Pre-process the obtained news text,using a text processing method based on fusion TF-IDF and LDA,and compare it with the traditional text processing algorithm,this method has a good performance in text processing.(3)The pre-processed text is classified into categories,the KNN algorithm,the Naive Bayes algorithm,and the decision tree algorithm are compared,and the improved Naive Bayes algorithm is selected as the text classification processing method.This improved method reduces common words.The difference in word frequency between the uncommon word and the uncommon word solves the problem caused by the consistent feature weight in the process of the classification of current news texts.And compare it with the traditional text classification method,the method shows good classification effect.(4)Apply the neural network algorithm model to the task of updating the current political hotspots,and use BP neural network algorithm to analyze the news text to carry out the hotspot update work.Experiments prove that the method performs automatic periodic update after the current political news hotspot analysis better performance.