Feature Fusion Recommendation Algorithm Based on Collaborative Filtering

In view of the different consumption behaviors and habits of users in different regions of a city, if the behavior data of users in the whole network are used for model training without distinction, some unique consumption habits of users in a certain region may be ignored. This paper proposes a feature fusion algorithm based on Collaborative Filtering. First, it makes full use of the mature regional data to learn a model with strong generalization ability, and then adds sparse data to conduct feature fusion, so as to recommend items that users with sparse features may buy more accurately. The validity of the collaborative filtering algorithm is proved by the order data of two different regions in a city.