Content-Based Recommendation Model in Micro-blogs Community

In this article, we present some novelty ideas to build a content-based recommendation system in micro-blogs community. We introduce a model to present users' preferences as a directed graph, named as "preference links", combining social relationships and people influences factors. Based on this model, we design an algorithm to collect recommendation candidates by visiting users' "preference links" and then generate a matrix to measure relevancies between content candidates and users' interests. A ranking function is proposed to rank these candidates based on the "relevancy matrix". We take the top items of the ranking results as the recommendation result. By implementing a prototype with these ideas in a real China micro-blogs community (Sina Weibo), our experiments show it can make personal recommendation with good accuracy.