This paper puts forward a recommendation algorithm of various factors based on the labels. Users can define the factors and sort them with priority according to their own requirements., This algorithm will select the resources according to the initial information of users, then update users’ groups and users’ preferences, then recommend the useful resources for users based on the similarity calculation between users and project and the correlation calculation of the project. The algorithm model adopts classification of combination to get the results, and reduces the complexity of similarity calculation. The algorithm was applied to the personalized learning platform of remote training platform in enterprise. The results show that this algorithm has much improved the recommendation effects of user's personalized learning resources.