doiSerbia

Text recommendation based on time series and multi-label information

Yin Yi (Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology; Key Laboratory of Data Storage System, Ministry of Education, Huazhong University of Science and Technology, Wuhan, Hubei, China + School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan, Hubei, China), yinyi@wust.edu.cn
Feng Dan (Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology; Key Laboratory of Data Storage System, Ministry of Education, Huazhong University of Science and Technology, Wuhan, Hubei, China), dfeng@hust.edu.cn
Shi Zhan (Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology; Key Laboratory of Data Storage System, Ministry of Education, Huazhong University of Science and Technology, Wuhan, Hubei, China), zshi@hust.edu.cn
Ouyang Lin (School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan, Hubei, China), ouyanglin@wust.edu.cn

One of the key functions of the method of text recommendation is to build a correlation analysis to all the text collection. At present, most of the text recommendation methods use the citation network, but less to consider the internal relations, which has become a challenge and an opportunity for the research of text recommendation. Therefore, we propose a new method to ameliorate the above problem based on the time series in this paper. We specify a certain text collection according to the interests of users and integrate the varied label values of the text, then we build the correlation coefficient between text and its related text with the differential analysis, finally the similarity degree of the text is calculated out by using the improved cosine similarity correlation matrix to promote a recommendation of similar text. Our experiments indicate that we are able to ensure the quality of text, with an improvement of accuracy by 8.63% as well as an improvement of recall rate by 5.25%.

Keywords: time series, label value, correlation coefficient, similarity degree