Research On The Influence Of Maximization Based On User 's Preference In Social Network

With the development of information technology and Internet technology, the websites with social networking features such as Facebook, micro-letters has achieved great success. The influence maximization problem which intended to mining the most influential collection of Top-k nodes in social networks is the key issue in the field of social network research. While a user may prefer to different topics in a different degree of social networks, and over time, the preference of the subject will also change. But in the previous works, the influence maximization problem is ignoring these factors and the excavations of users are the most influential global mode users. The accuracy of conventional algorithms will be greatly reduced if we want to find the most influential users under specific topic in current time.Against this background, we propose the user-preference-changing-based influence maximization problem and build a UCP_IC Influence propagation model which considers user preference change. So as to solve the problem of user preference change, the model design an exponential function to measure the user current preference based on Ebbinghaus forgetting curve. Furthermore, in order to associate user’s preference with user’s activation probability, We consider the frequency of contact with the user’s preferences in a particular topic, and the use of association rules method will link the two as the activation probability between users. Based on the model, we propose a new GAUCP algorithm to find the most influential users on a given topic in the current time. In consideration of the user’s current preferences, the algorithm use climbing greedy algorithm to mining customers. In particular topic, it can achieve better accuracy. The algorithm can be obtained about63%accuracy guarantee base on the submodular of Influence propagation model, and it can use CELF algorithm to optimize the computational efficiency.Finally, based on academic databases DBLP related experiments, GAUCP can find the most influential set of user under particular topic in current time.