Research On Topic Recognition And Trend Prediction In Technical Q&A Community

In the Internet era,knowledge sharing has become a new way of knowledge interaction.Netizens are more and more inclined to ask or answer questions in the Q&A community to exchange knowledge with each other,which benefits both sides.Different users excel in different areas,therefore,for the Q&A community,it is very important to make the users find questions about related topics accurately and quickly,or understand the development trend of a topic obviously,and improve the users’ experience in the Q&A community.Although traditional LDA model and the single time series prediction method also can meet the above requirements,there are shortcomings such as topic recognition result is not accurate,the error rate of trend prediction is large,so it is difficult to meet the needs of users for topic recognition and trend prediction of Q&A community.To solve the above problems,this paper proposes a method to identify hot topics and predict trends in technical Q&A community based on improved LDA and contingency model.Firstly,using the traditional LDA model to extract topics and keywords from the training dataset,and the results are used for word vectorization,keyword similarity calculation and weight filtering,then filter topics and keywords again through the Text Rank algorithm.Secondly,constructing times series according to the topic category of the Q&A dataset,then predicting times series by ARIMA and Holt-Winters,calculating the MAE of these two predictions and combining them with the least square method to obtain the weight values of the contingency combination model and the prediction results of the combination model.By comparing the Precision,Recall and F1-Score values of the LDA model and the improved LDA model,as well as the MAE,RMSE and MAPE values of the prediction results of ARIMA model,Holt-Winters model and contingency combination model.The comparison results verify the superiority of the improved LDA model and the contingency combination model proposed in this paper.Finally,the algorithm model proposed in this paper is applied to the CSDN Q&A community,it can obtain the popular topic and development trend of the Q&A community,which proves that the algorithm model in this paper has good practical application.