改进协同过滤算法在服装个性化推荐的研究

Clothing Personalized Recommendation Research Based on Improved Collaborative Filtering Algorithm

基于项目的协同过滤算法大量应用在电子商务推荐系统上,传统的基于项目的协同过滤算法对物品的相似度计算方式单一,难以排除对用户评分标准、项目流行度、项目品质的影响.利用改进因子(包括用户评分权重、项目优劣性、项目共同用户数)方式,来改进相似度的计算,应用于服装推荐领域,降低了绝对平均误差,使得推荐效果更加准确.

Item-based collaborative filtering algorithms are applied to recommendation system.But tranditional item-based collaborative filtering algorithms are simple in similarity calculation so that it’s hard to rule out the influence of difference of user rating criteria,item popularity and item quality.We use improvement factors including user rating weight,item quality,number of public user in improving calculation of similarity,which can apply to clothing recommendation.It reduces mean absolute error and makes the result of recommendation more accurate.