多属性协同过滤推荐在物流配送服务平台的应用

为了帮助用户选择尽可能满足其个性化偏好的物流配送服务，结合配送服务的多属性评分特点，本文构建了基于多属性协同过滤的推荐算法，对传统协同过滤算法进行了延伸与改进，首先预测目标用户对候选服务各属性的评分值，通过引入服务的个性化特征因子减小热门服务对用户相似度计算的误差，考虑到用户的服务属性评分存在波动性，使用信息熵将用户历史评分均值与协同过滤得到的预测值相结合进行修正，然后基于同一用户对不同属性评分波动性的差异，计算得到用户对服务所有属性的评分预测权重，将各属性的评分预测值与对应权重加权求和进行服务推荐.对配送服务交易的评分数据样本进行实验验证，在准确率和平均绝对误差指标上有较好的表现，将算法应用于物流配送服务平台，构建推荐系统，能够提高平台个性化服务能力.

Application of Multi-Criteria Collaborative Filtering Recommendation in Logistics Distribution Service Platform

In order to help users to choose the logistics distribution service that meets their personalized preferences as much as possible, combined with the multi-criteria rating characteristics of the distribution service, this study constructs a recommendation algorithm based on multi-criteria rating collaborative filtering and extends and improves the traditional collaborative filtering algorithm. The target user''s rating of each criterion of the candidate service is reduced by the introduction of a personalized feature of the service to reduce the error of the hot service to user similarity calculation. Considering that the user''s service criterion rating is fluctuating, the information entropy is used to average the user''s history rating. It is combined with the predictive value obtained through collaborative filtering. Then based on the difference in the volatility of different criteria of the same user, the score of the user''s rating for all criteria of the service is calculated, and the predicted rating value of each criterion is weighted with the corresponding weight to make service recommendations. The experimental data of the rating data sample of the distribution service transaction is verified by experiments. The accuracy and average absolute error index have better performance. The algorithm is applied to the logistics and distribution service platform to build a recommendation system, which can improve the platform''s personalized service capabilities.