Management and Monitoring of Multi-Behavior Recommendation Systems Using Graph Convolutional Neural Networks

Different recommendation algorithms, which often use only a single type of user-item engagement, are plagued by imbalanced datasets and cold start problems. Multi-behavior recommendations, which takes advantage of a variety of customer interaction including click and favorites, can be a good option. Early attempts at multi-behavior suggestion tried to consider the varying levels of effect each behavior has on the target behavior. They also disregard the meanings of behaviors, which are implicit in multi-behavior information. Because of these two flaws, the information isn't being completely utilized to improve suggestion performance on the specific behavior. In this paper, we take a novel response to the situation by creating a unified network to capture multi-behavior information and displaying the MBGCNNN model (Multi-Behavior Graph Convolutional Neural Network). MBGCNN may effectively overcome the constraints of prior studies by learning behavior intensity via the user-item dissemination level and collecting behavior interpretation via the items dissemination level. Practical derives from various data sets back up our model's order to leverage multi-behavior data. On real methods, our approach beats the average background by 25.02 percent and 6.51 percent, respectively. Additional research on cold-start consumers supports the viability of our suggested approach.