基于复杂网络及神经网络挖掘用户兴趣的方法

摘要

Providing personalized service according to the user’ s interest is the most effective solution to improve the commercial value. Aiming at the problem of mining user’ s interest method from user behavior,a method of constructing complex network based on time se-ries and neural network clustering is proposed,which is based on the user’ s software. In the calculation of user interest in software,the u-sing time and adjacent nodes are considered including node degree,betweenness and clustering coefficient to determine the node impor-tance for mining user for the degree of interest for the software,forming of interest community. Then the neural network is used to cluster the software in the user interest community. The experiments show that this method can be more accurate than other methods to mine the user’ s interest,and the accuracy rate and recall rate of the algorithm is improved.%按照用户的兴趣提供个性化服务是提高企业商业价值最有效的方案。针对目前从用户行为中挖掘用户兴趣方法的不足，提出一种依据用户使用软件的时间序列构建复杂网络及依据神经网络聚类挖掘用户兴趣软件的方法。在计算用户对于软件的兴趣度时，综合考虑用户使用软件的时长以及复杂网络中相邻节点的贡献度，包括节点的度、节点介数、聚集系数来判断节点的重要性，挖掘用户对于软件的兴趣度，形成软件兴趣社区。再利用神经网络算法对用户兴趣社区中的软件进行聚类，形成用户的兴趣软件集。实验结果表明，该方法能够较准确地挖掘用户感兴趣的软件集，并且在精确率和召回率上较其他方法有一定的提高。