Abstract: Internet technology and online social networks have developed rapidly, which enables people to randomly express their opinions, ideas, emotional exchanges and economic exchanges. Inferring social networks is made possible through the observation data exchanged by people on the Internet. Through the analysis of ConNIe (Convex Network Inference) algorithm, this paper researches the effects of sparse parameter, propagation time distribution model and its parameters on the inference results of the algorithm. According to the analysis, a social network inference framework based on ConNIe algorithm is proposed. Combining the perceptron and particle swarm optimization algorithm, the ConNIe algorithm inference becomes a complete system. The research in this paper has a widely practical value in the fields of social public opinion control and marketing.

Key words: social network, network inference, ConNIe, perceptron algorithm, particle swarm optimization