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.