Research Of Influence Maximization Based On Topic In Academic Network

Viral marketing, a very effective marketing strategy of conducting product promotionsthrough word-of-mouth, which spread social influence among individuals’ cycles offriends, families, or co-workers. Motivated by this background, influence maximizationproblem has become a focus in academic, the research community has recently studied thealgorithmic aspects of maximizing influence in social networks. Influence maximization isthe problem of finding a small subset of nodes (seed nodes) in a social network that couldmaximize the spread of influence. Finding out the most influential classic papers andresearch leader in the academic network is important to promote the scientific researchcooperation and guide the scientific research work.A academic network is modeled as a weighted direct graph with vertices representingpapers and edges representing citation relationship between two papers. Influence spreadin the network through a influence spread model. Most of these works are based on thetwo basic influence spread models, namely linear threshold model (LT model) andindependent cascade model(IC model), and their extensions. According to the topic model,a improvement spread model of the LT model which named Linear Threshold based onTopic model (T_LT model) is proposed. By summarizing the characteristics of theclassical greedy and heuristic algorithms, a new influence maximization algorithm basedon topic model which named TIM algorithm is put forward. TIM algorithm consideringboth the structure characteristic and propagation characteristics of the academic network,first heuristically choose the initial nodes with the biggest influence value based on topicclustering, and then greedily to select the rest of the most influential nodes.Finally, the experiment employs the academic citation network data sets of theArnetminer platform, we contrast our method with classical greedy KKT and MaxDegreeinfluence maximization algorithms from aspects of both effectiveness and efficiency.Experimental results show that: TIM algorithm can greatly reduce the timecomplexity under the approximate effect of the KKT algorithm, and TIM algorithmimproves the effect compared with the MaxDegree algorithm. TIM algorithm is aneffective algorithm for the influence maximization problem.