A recommendation algorithm model based on the fusion of group behavior and temporal feature

With the rapid development of the Internet and the exponential growth of network information resources, how to help users quickly obtain valuable information has become a major challenge. As the key technology of this problem, recommendation algorithm has become an important tool in the information age. However, most existing recommendation systems, such as traditional deep learning recommendation algorithms, still have problems such as cold start of new users and improper handling of dynamic changes in user behavior. To solve these problems, this paper proposes a fusion Group behavior characteristics and time series characteristics of the recommendation algorithm model Group-Temporal Recommendation Fusion (GTRF). Based on deep learning algorithm, the model extracts group behavioral characteristics by using graph attention network to deal with the problem of cold start, obtains time series characteristics by using gated convolution and GRU to deal with the problem of dynamic changes in user behavior, and finally uses cross-attention mechanism to effectively integrate these two types of features to achieve high-precision prediction of recommended content. GTRF not only solves the problems such as cold start of new users and dynamic change of user behavior, but also significantly improves the comprehensive performance of recommendation algorithm, showing the indispensable and great potential of recommendation algorithm in information retrieval.