Research Of Recommendation Method Base On Community User

Recommendation technology can enhance the user experience in the information load today. Its main task is to identify user intention, predict user behavior and help users to filter information. Micro-blog environment as an example, many scholars introduce the characteristics of the recommendation process is less, and the comprehensive consideration is not complete. At the same time, the importance of community user relationship and user behavior is also reflected in the recommendation of computing research. In this paper, the user interest change and social information of the user were studied respectively.For user interest, the LDA topic model is applied to micro-blog recommended. Micro-blog content as the basis of analysis. First of all, a large number of micro-blog content filtering, word segmentation and stop word processing. After, treatment of each corresponding words by micro-blog LDA topic model. The second step, according to each micro-blog has the theme of clustering to build interest classes. Each interest class includes a number of topics. The third step, the user’s release of micro-blog and users to participate in the interaction of micro-blog’s theme in accordance with the inclusion of interest in the relationship between the establishment of a model describing the user’s interest. The third step, the user’s micro-blog and related topics in accordance with the interests of the relationship, the establishment of a model of user interest. The fourth step, according to the time axis, calculate the interest model difference between the two moments, get the interest change, finds a tendency. According to the tendency of interest included in micro-blog, this paper designed two methods to carry out micro-blog recommended. At last, two methods are compared in the experiment.In the second research, this paper applies the social information of the users in the collaborative filtering recommendation strategy. The first step, according to the user’s personal description, social, behavior and interest information, the establishment of social attribute fusion model. The second step, through the model information to find the current user characteristics and the performance of similar neighbors. The third step, according to the selected neighbor users, using collaborative filtering recommendation method, the content of the corresponding project recommendation. In the selection of neighbor users, this article compare with based on social information for the selection of neighbor users and based on the similarity score of neighbor users. The experiment proves that the latter is better than the former. In the items of recommendation, we improved the numerical value of the score vector, and compared with the result of the improved recommendation, it shows that the improved scheme is effective. In the process of recommendation, we improved the numerical value of the score vector, and compared with the result of the improved recommendation, it shows that the improved scheme is effective.Innovative points are as follows:1 The LDA topic model will be used in the micro-blog environment, Create an interesting class layer on the topic model. User interest can be recorded by means of interest vector. The choice of strategic interest in the strategy can reduce the number of micro-blog search. At the same time, according to the change of user’s interest, two recommended methods are designed.2 Integrated social information content for users to quantify the description. Combining the social information and the collaborative filtering recommendation method. At the same time, to ease the problem of the users in collaborative filtering, the score vector is sparse.