Abstract: To improve the accuracy of detecting hot topic from social network text information, a hot topic identification algorithm based on user relevance analysis is raised. The algorithm considers both the frequency change rate of feature word and the authority of users. The frequency change rate of feature word is elevated using EMA and MACD indicators and the authority of users is calculated by creating user relevance graph. We use a method based on HITS algorithm to calculate the hot value of topic by combining feature frequency change rate data with user authority data together. According to the result of the experiment, the hot topic identification algorithm based on user relevance analysis can raise the accuracy of hot topic identification.

Key words:  hot topic detection, user authority calculation, feature change rate calculation