A HMM_based hot topic lifecycle prediction model

Author

Liu, Ruifang ; Wang, Jun ; Zhang, Meng

Author_Institution

Sch. of Inf. & Commun. Eng., Beijing Univ. of Posts & Telecommun., Beijing, China

fYear

2011

fDate

15-17 April 2011

Firstpage

2881

Lastpage

2884

Abstract

Web documents can be clustered into topics with topic detection and tracking(TDT) technologies. With the topics´ data collected by TDT system, it is found that the lifecycle of topics has 4 stages. In the paper a HMM-based state prediction model for topics is proposed. Some topics with similar lifecycles share a same model, several models are trained with history data of topics, these models are used for new topic state prediction. Experiment results show the performance of the Forward Probability Prediction Algorithm, and the comparison with other method is analyzed. It will be useful for the department who concern the hot topic monitoring.

Keywords

document handling; pattern clustering; HMM_based hot topic lifecycle prediction model; TDT; forward probability prediction algorithm; topic detection and tracking; web documents; Data models; Event detection; Fuzzy systems; Hidden Markov models; Internet; Monitoring; Predictive models; HMM; hot topic prediction; prediction model; topic detection and tracking;

fLanguage

English

Publisher

ieee

Conference_Titel

Electric Information and Control Engineering (ICEICE), 2011 International Conference on

Conference_Location

Wuhan

Print_ISBN

978-1-4244-8036-4

Type

conf

DOI

10.1109/ICEICE.2011.5777416

Filename

5777416