Design And Implementation Of Online Friend Recommendation System Based On Weibo

Living under the social network environment, people’s social approach has largely been shifted to online. Normally, a higher quality and broader social relationship is often a key to promote our own value and self-development. As online social networking is actually a continuation and supplement of real life social contact, people are willing to expand their social circle in social network as they do in real social life. However, friend recommendation can only be conducted by computer except for manually added. Therefore, friend recommendation is a topic worth studying, and the importance is self-evident.Aiming at this problem, the existing methods are mainly concerning on the mining of recommendation factors in depth and the strategy of combining some of these factors. Friend recommendation cannot meet the needs of the majority if takes under the consideration of any single aspect as everyone has their own preference. Therefore, this thesis puts forward comprehensive recommendation method based on the fusion of many recommendation factors, including user’s interests, social relationships and geographic location information. It is worth mentioning that this fusion method is different from the normal fusion method which works by filtering the recommendation list with some other recommendation factors. The method used in this thesis, which makes each recommendation factor works independently with others in the process of calculation, is based on various factors. Friend recommendation based on the interests of users uses their weibo context for interest mining. During the recommendation process, we use the improved cosine similarity algorithm based on the characters of weibo, and improves the clustering effect by means of quadratic clustering. Friend recommendation based on social relation uses their friend relationship and users’ personal profiles. Friend recommendation based on geographical location information takes advantage of users’ POI(Point Of Interest) information, and makes recommendation based on the distance between users. Finally, this article designed and implemented this kind of recommendation system on android operation system. The android client has the feature of simple design, utility functions and the fool operation, etc.Experiments based on this recommendation system have proven that this algorithm is better than others. In consequence, this system not only can meet the demand of the majority, but has practical value as well by taking advantage of its ability of flexible configuration.