Research On Information Recommender Systems Based On Context

With the advent of the era of big data, "information overload" problem becomes more serious, in order to make a good match between supply and demand issues of the information providers and information users, the Internet intelligence products-recommender systems come into being.This paper which does a research on information recommender systems based on context-aware is the traditional recommender systems’ expansion and extension based on them. The paper uses respectively the two common ways of attitude scale made by the semantic difference method and text mining from unstructured product reviews to extract the user emotional preference, then information recommendation system model based on emotional factors in this context is presented, which considers the emotional factors into recommended model, making the recommender system bring greater accuracy and better customer satisfaction. The main work is as follows:1. Research on CARS based on Sensory Semantic context-aware. A special preferences of the user-oriented Kansei recommendation method which considering appraisement new product recommended problem is proposed. In the method, firstly, semantic differential(SD) method and linguistic quantifier are used to design a q uestionnaire for gathering Kansei data of new product schemes and the possibility distribution of feeling target in different preference orders. Then, a recommended utility function considering emotional needs of the user-oriented is presented. And the use of hybrid weighted average operator is calculated and gathered for evaluation information of each program. Finally, the alternatives are ranking according with the utility function value of each program. Especially when given special users’ demands, the matching between new product alternatives and user preferences can be calculated. The analysis for a numerical example illustrates the feasibility and validity of the proposed approach.2.Research on information recommender system based user context aware. Exploration of mining user context or emotional context information from unstructured text in reviews, better results for recommendation system after adding context into recommendation model are produced. Firstly, through mining context from the user requirement description, the current context of user is determined. Among them, the part of context reasoning module is bycrawling Jingdong book reviews dataset and using Gibbs sampling to train L-LDA model to get a classifier, then select the topic which the highest probability category is assigned to as the current context of user in the description of user needs. Meanwhile, using standard k-nearest neighbor algorithm or collaborative filtering algorithm to predict the most suitable item i to user u, the traditional score prediction value is got. Given the context, matching degree between user u and item i is calculated, then the most consistent item i with the current user context is chosen. Combined with the traditional prediction score weighted into matching degree on the state of context, the<u, i> utility value is calculated by the two part of the results. According to the value, each item will be sorted, and the optimal recommendation will be got. Finally, test data climbed Jingdong books is set to verify the feasibility and effectiveness of the method. This paper focuses on making an exploration and research on the emotional context of the factors that impact on the conduct of the recommendation system, and respectively uses two common ways to extract user preferences and establish a corresponding recommended utility function including emotional context. Finally, the validity of the model is done.