基于用户属性和活跃性的协同过滤推荐算法

AN RECOMMENDATION ALGORITHM BASED ON ATTRIBUTES AND ACTIVITIES

随着大数据时代的到来,信息过载问题日益凸显,个性化的推荐服务是解决该问题的有效手段之一,因为其简单、高效的特点,越来越受到人们的重视.协同过滤是个性化推荐的常用手段,协同过滤推荐算法通过研究用户的喜好,实现从海量数据资源中为用户推荐其感兴趣的内容,在很多领域中都得到了广泛应用.但是,冷启动和数据稀疏依然是其面临的难题,在某些领域中,出现推荐算法效率偏低,推荐准确度下降问题,导致用户满意度偏低.针对这个问题,本文提出了用户属性相似度概念及移动图书馆中的活跃相似度,并融入了基于内容过滤的算法思想,提出了一种改进的协同过滤推荐算法.实验结果表明:改进的算法能有效提高推荐准确性,并在一定程度上缓解了冷启动的问题.

With the coming age of big data, the problem of information overload is getting more and more prominent. As one of the effective means to solve this problem, personalized recommendation service is receiving increasing attention due to its simplicity and efficiency, of which collaborative filtering is a common method. By studying the users＇ preference, collaborative filtering recommendation algorithm makes it possible to recommend users the content they are interested in from the massive data resources, and as a result,it is widely used in many fields. However, cold start - up and data sparseness are still the conundrum. In some areas, low users＇ satisfaction is caused by low efficiency and declining accuracy in recommendation algorithm. To solve this problem, this paper proposes the concept of users＇ attribute similarity and activity similarity in mobile library, and presents an improved collaborative filtering recommendation algorithm integrating with algorithm thinking based on content filtering. The experimental results showed that the improved algorithm can effectively improve the recommendation accuracy, and to certain extent, mitigate the problem of cold start - up.