Abstract: Studies user collaborative filtering control recommendation based on behavior curve, and creatively proposes an optimized UBC-based CF algorithm: Collaborative Filtering Based on User Of Behavior Curve. This algorithm uses more fine-grained behavior curves to represent users and the similarity of behavior curves to calculate the user's previous similarity, so as to provide customers with a more precise air conditioning control recommendation service. The method mainly includes the following steps: Firstly, extract the user's historical operation behavior and generate a behavior curve to represent the user; Then, calculate the Jaccard similarity of each two behavior curves to evaluate the similarity between each two users; Next, for the recommended user, select a certain number (K) of neighbors closest to the user and combine with the current control parameters to generate the user's parameter recommendation; Finally, several experiments are conducted and the effect is verified. It can be seen from the experimental results that the accuracy of this method can reach 91% and the effect is optimized.

Key words: User behavior, Real-time recommendation, Automatic control