融合位置信息和物品流行度的协同过滤算法

Collaborative filtering algorithm combining location information and item popularity

针对绝大多数用户消费习惯对地理位置的敏感性,以及推荐过程中的"长尾效应",提出融合位置信息和物品流行度的协同过滤算法。对传统的协同过滤算法作出2点改进:第一,将用户兴趣偏好与位置偏好相结合,提出一种新的基于地理位置的用户相似度计算方法;第二,在预测评分时,引入物品流行度权重,合理地调整流行物品和长尾物品的推荐期望值。使用Foursquare数据集作为实验数据集,与相关算法进行对比实验。结果表明,改进算法能有效提高推荐的精度和推荐结果的多样性。

Aiming at the sensitivity of most users’ consumption habits to the geographical location and the "long tail effect" in the recommendation process, a collaborative filtering algorithm is proposed combining the location information and the item popularity. In this study, two improvements are made to the conventional collaborative filtering algorithm. At first, a new calculating method of user similarity based on the geographic location is proposed by combining the preference of user interest and the position preference. Then, in predicting, the recommended expectations of the popular items and the long-tailed items are reasonably adjusted by introducing the popularity of goods weight. The Foursquare data set is used as the experimental data set in this work, and experiments are compared with related algorithms. The experimental results show that the improved algorithm can effectively improve the precision and the diversity of the recommended results.