一种改进的协同过滤的商品推荐方法

An Improved Co-filtered Goods Recommendation Method

传统的电子商务平台无法实现对用户进行个性化优质商品的推荐,大量的商品信息无法被充分应用在帮助客户选择商品上。针对上述问题,提出了改进的协同过滤算法(new Pearson collaborative filtering,NP-CF)为用户进行个性化的电子商品推荐。该算法弱化了活跃用户的对商品相似度的贡献程度并且利用标准差的计算降低电子商品本身质量对相似度的影响,将两者计算获得的系数与皮尔逊关系数相结合,从而计算出更加准确的用户相似度,再利用相似度值计算商品的推荐值并且通过加权评价公式对该值进行加权处理。最后在真实的数据集MovieLens和人工数据集Mobile-Data上对该算法进行实验测试,且与传统的基于用户信息的协同过滤算法(user collaborative filtering,User-CF)进行比对,该算法(NP-CF)整体上优化了推荐结果并且提高了推荐的准确率。

Traditional e-commerce platforms are unable to realize personalized high-quality goods recommendations for users,causing users to face a large amount of goods information,but the utilization rate of information is gradually decreasing.In response to the above problems,we propose an improved collaborative filtering algorithm(new Pearson collaborative filtering,NP-CF)to provide users with personalized electronic goods recommendations.The algorithm weakens the contribution of active users to goods similarity and uses the calculation of standard deviation to reduce the impact of the quality of electronic goods on similarity.The calculated coefficients of the two with the Pearson relationship number is combined to calculate a more accurate user similarity,then the similarity value is used to calculate the recommended value of the goods and weight the value through a weighted evaluation formula.Finally,the proposed algorithm is tested on the real data set called MovieLens and the artificial data set named Mobile-Data and compared with the traditional user collaborative filtering based on user information(user-CF).The proposed NP-CF optimizes the recommendation results as a whole and improves the accuracy.