Topic Popularity Prediction Method Based on Similarity and Co-occurrence

Abstract

The invention belongs to the technical field of social network big data, and particularly relates to a topic popularity prediction method based on similarity relation and co-occurrence relation. The method includes main steps: (S1), inputting a document set, and analyzing through a LDA modeling to acquire a topic set, vocabulary distribution of each topic and topic distribution of each document; (S2), calculating similarity vector of topics; (S3), calculating co-occurrence relation vector of the topics; (S4), defining popularity, and calculating topic popularity; (S5), calculating popularity of the topics based on the similarity relation; (S6), calculating popularity of the topics based on the co-occurrence relation; (S7), according to calculation results of the step (S5) and the step (S6), respectively calculating probability values of topics under four types of popularity, and selecting the type with the highest probability value as final popularity of the topics. The method applies topic popularity prediction in social network big data and realizes short-term prediction of topic popularity.

Description

technical field [0001] The invention belongs to the technical field of social network big data, and in particular relates to a topic popularity prediction method based on a similarity relationship and a co-occurrence relationship. Background technique [0002] As one of the most popular forms of online social networking today, Weibo has developed rapidly in recent years. It is an interactive and extremely fast-spreading platform, and its spreading speed is even faster than that of general social networks and media. With the emergence of microblog, many news events and hot topics can be spread quickly and widely on the microblog website, and it has become an important platform for netizens to obtain information, share information, and make friends. The topics published by users will receive different degrees of attention according to their authors and contents. In order to better recommend valuable and popular topics to users and to sort the popularity of topics reasonably, i...

AI Technical Summary

Problems solved by technology

[0003] references [2] An algorithm suitable for long-term prediction is proposed. The basic idea is to treat the predicted topic, assuming its historical heat curve, and predict the future trend of the topic according to the heat curve of other topics that are most similar to its historical heat curve. The shortcomings of this method is unable to deal with the problem of short-term forecasting

Although this method uses the LDA model to make up for the problem of the vector space model, it does not consider the co-occurrence relationship between topics. There is a certain logical relationship between co-occurrence topics, which plays an important role in predicting topic popularity.

Images