计算机与数字工程2017，Vol.45Issue(10)：1980-1985,6.DOI:10.3969/j.issn.1672-9722.2017.10.019
一种基于RLDA主题模型的特征提取方法
Feature Extraction Method Based on RLDA Topic Model
冯新淇 1张琨 1任奕豪 1谢彬 1赵静1
作者信息
- 1. 南京理工大学计算机科学与工程学院 南京210094
- 折叠
摘要
Abstract
In this paper,to accurately mining micro-blog user interest,the data concerning original,reposted and liked mi-cro-blog content as well as the ranking of all these micro-blogs are collected and analyzed.So the accurate description information of micro-blog users'interests is obtained. Then based on the LDA model,we proposed a modified topic feature extraction model named as Ranking LDA is proposed.In comparison to LDA model,RLDA model includs a new concept-Micro-blog popularity rank-ing to improve the mining accuracy of the micro-blog users'interests.In the process of modeling the RLDA topic model,the con-cepts of hyper-hyper parameters is introduced.Hyper parameters are sampled from dirichlet distribution.Experiments suggest that, compared with the LDA model,RLDA model achieves quite a great promotion on the accuracy of interest mining for micro-blog users.关键词
兴趣挖掘/微博热度排行/RLDA模型/特征提取/超超参数Key words
interests mining/Micro-blog popularity ranking/Ranking Latent Dirichlet Allocation model/feature extrac-tion/hyper-hyper parameters分类
信息技术与安全科学引用本文复制引用
冯新淇,张琨,任奕豪,谢彬,赵静..一种基于RLDA主题模型的特征提取方法[J].计算机与数字工程,2017,45(10):1980-1985,6.
