Study of the clustering result based on user behavior feedback

2017 IEEE 2nd International Conference on Cloud Computing and Big Data Analysis (ICCCBDA)(2017)

引用 1|浏览14
Currently, similarity search is usually based on user's keywords, providing the sorted result to the user according to the similarity search algorithms. Sometimes, some results ranking ahead are not what the user really need. Therefore, we put forward the algorithm called KPathsim, which return the results based on users' behavior. Firstly, KPathsim algorithm return the sorted result based on the similarity calculation formula of given paths (e.g., two objects are similar if they are linked by many paths in the network). Secondly, it use the user's choice as clustering seeds to calculate the similarity and return the sorted result. Finally, we use the NMI to calculate the clustering accuracy. Experiment shows that it significantly improves the search accuracy Compare to the algorithms of Pathsim and PCRW. The clustering accuracy of NMI can improve according to some certain given paths. It not only improve the user's satisfaction, but also a kind of method of similarity search based on user's feedback compared with other pure clustering algorithms, adding the meta-path semantic message in the heterogeneous information network.
heterogeneous information network,meta-path,link path,feedback,clustering
AI 理解论文