A method of knowledge item recommendation based on Skill-LFM

At present, the users of knowledge base mainly get the required knowledge items through search, which relies on the search engine to solve the information overload problem. It is inefficient for real-time online services, and has no integrity and continuity of offline knowledge learning. Therefore, it is proposed that knowledge items should be actively recommended to users by the knowledge base system according to their level of skills, to improve the efficiency of decision making, and also to help users establish a complete knowledge learning system. A collaborative filtering recommendation method is proposed to predict every user's preference on knowledge items, based on the historical behavior of a user on the knowledge items, and the knowledge learning ability of this user. This method combines latent factor model with skill, named Skill-LFM, where the difficulties of knowledge items are taken as potential factors, and users' ability level is considered to give personalized recommendations. Tested on the data from a call center knowledge base, the proposed Skill-LFM outperforms the baseline latent factor model in terms of lower RMSE. Considering the characteristics of the application domain and the historical behavior data of the knowledge base, this paper demonstrates the possibility of further improving knowledge item recommendation through integrating user and knowledge item context information.