Research On Recommended Incentive Strategy Based On Game Theory In Social Network

With the advance of the time, people's daily life can't be separated from social network, which makes users' lives more convenient. The way to acquire knowledge is also becoming diversified, and it would save time and effort by shopping online. However, the network followed by exponential growth of information brings about the "information overload", so that the user would not be able to search for the interested information from a lot of information, which would increase the time in user's experience from network. Therefore, the recommendation system for filtering information comes into being. It appears greatly to reduce the time of search information, and provide a better experience for users. Recommendation mainly relies on the trust relationship between users in trust-based recommendation, but the trust data is sparse, and the coverage of recommendation system is low.In this paper, the comment behavior of the user's is studied to improve the problem of data sparsely in the user ratings and trust matrix, the collusion from the recommendation in groups of users, and the new user's recommendation. According to similar interests of the trust users could be regard trust as a weighted to predict rating, and then combined with the rating of the project is able to improve the accuracy of recommendation. The rating motivation of the users comes from the user's tendency of benefit. In order to get more benefits, users would be give effective ratings. Therefore, the incentive strategy is proposed to urge customers to select more rational evaluation. In short, The main works of this paper are as follows:1. An improved recommendation based on trust network is proposed in this paper. According to the inconsistency of user's comment criterion, the original trust network construction method in trust-based recommendation is studied to build a new trust model in this paper. The new trust model including the explicit trust which is set by users, the implicit trust which is consist of users rating and the feedback trust. In this paper, the data from Epinion website is used to simulate the traditional recommendation's and the proposed recommendation's performance in accuracy and coverage. The experiment results show that compared with the original trust-based recommendation, the trust model that this paper proposed played better in accuracy under certain conditions.2. The incentive strategy of user reviews different evaluation criterions is proposed in this paper. The excitation function is constructed for the motivation of users' comments and based on game theory. Excitation function consists of two parts: a fixed excitation based on other users rating and a dynamic excitation based on feedback trust. Finally, we analyze the probability that the user select the comment action based on game theory. The analysis shows that in order to obtain more benefits, the rational user would select the comment strategy, and collusive user would not choose the comment strategy as time goes. Therefore, the excitation function would facilitate the commenting behaviors of the users which resulting in more ratings3. A new method is proposed to solve the new users recommended which based on the user's authority. The users which has max expected revenue is set to the authority user. Preliminary simulation results based on Epinion dataset showed that the algorithm for new user's recommendation has a better performance.