Research On Detection Of Emerging Hot Topics On Micro-Blogging

Micro-blogging, a new social network platform which provides users of information gaining, sharing and spreading based on users’ social relations, appears along with the development of technology and the actual needs of users. At present, researches on information mining, micro-blogging marketing of commercial goods and sentiment analysis of public opinion emerge in endlessly. All these works aim at mining valuable information from the huge data repository, in order to help political-legal departments for social supervision, enterprises for marketing, and the development of smart city.The major research of this paper is mining emerging hot topic in the specified period based on users and documents data from micro-blogging. The main process is as follows. Firstly, calculating authority value of user based on social network of users according to Page Rank algorithm. Then, modeling life cycle of keyword, calculating its nutrition based on user’s authority value and weight of keyword, and then converting nutrition value to energy value in the specified period to mining emerging hot keywords. At last, creating a topic graph based on semantic relationships of keywords, and then taking an emerging hot keyword as the center for semantic, finding out a set of semantic relevant keywords in the same period to form emerging hot topics. The energy of topic is calculated by the energy values of all relevant keywords. This paper conducted experiment on real dataset and tested the results and the performance of the algorithm.The contributions of this paper are:1. Proposing a keyword-based energy computing method and a user authority calculating method based on Page Rank algorithm. In this paper, topic is split into a set of semantic relevant keywords. Life cycle modeling, nutrition value and energy value calculating are taking keyword as the unit and introducing user authority.2. Proposing a user-based TF*PDF method to compute weight of keyword. Dividing micro-blogging documents into groups according to its publisher and weight of keyword into relative weight and absolute weight. This means different users affect keywords differently.3. Creating a keyword-based topic graph, and finding out a set of keywords semantic related with the emerging hot keyword. Also, proposing a new topic energy calculating method.