Recommendation Method Based On User Interest Model And Community Detection

Zhihu is one of the most widely used socialized question-answer networks in China.Its rich data resource provides a huge opportunity for scientific research.The development of “knowledge e-commerce” and “knowledge payment” has become a trend.Knowledge can be treated as a virtual commodity and acquired by the user.The platform can also recommend questions and answers to the user according to their interests.At the same time,Zhihu is a social network.Its question and answer mode reflects a social feature.The interaction on the same question between users reflects their latent same interests.Based on the topic tags of Zhihu,the thesis proposed a user interest model which fused question topics and user topics.To validate the model,the repetition rate between question topics set and user topics set was calculated based on real data.Its value spans from 60% to 80%,which shows the model is valid.The sociability of Zhihu is based on the interaction around the question.Through testing and analyzing the statistical features of question based user networks and concern-based user networks,“small world” and clustering characteristics of question based user networks were validated.Furthermore,the advantages for implying user interests of question based user networks were analyzed and the construction method for user interest networks was proposed.The theory of complex network community discovery studies the strong and weak relationship in the network.In the social network,people with the same characteristics have a strong relationship structure.In Zhihu,there are some central nodes known as “opinion leader” that play the important role in each community.So the thesis proposed a GN algorithm with central nodes to detect the community.Based on the real data,the algorithm was tested to be efficiency and practical.This thesis also designed the crawler to obtain the real data for above-mentioned experiment.