Research On Technologies Of Hot Topic Detection And Topic Trend Prediction

With the rapid development of Internet technology,the Internet has become an indispensable part for people's work,study and life.Internet media has become the veritable “the fourth media”.The emergence of a large amount of network news has broadened people's horizon,enriched people's life,but they brings us the information overloaded problem at the same time.The spread transmission of rumors and false information has brought new challenges for public opinion monitoring.In order to help people solving the problem of information overloaded,provide technical support for public network opinion monitoring,this paper does meticulous research on technologies of hot topic detection and topic trend prediction.In this paper,the main work can mainly be divided into three parts.First,this paper proposes a clustering algorithm called “Two times single-pass clustering based on segmented time line” against the disadvantages of the traditional single-pass.Time distance factor is used to segment online documents into sections and the traditional single-pass is applied to complete a two-part incremental clustering,which improves the accuracy of topic detection.Secondly,life circle model of biological theory is applied for modeling the energy value of topics.Because of the difference of clustering algorithm,the combination method of traditional life circle model and single-pass is not suitable for the algorithm in this paper.The idea of doing clustering first and then calculating energy value of topics is adopted,the time point is drawn back to the start point of the time distance after clustering then energy value of topics is updated.Thirdly,this paper proposes a new way for topic trend prediction.Based on the constructed life circle model of topics,topic trend of energy values is traced using moving average model and topic trend prediction is finished combining with the calculating process of growth rate.