Deep interaction network based CTR prediction model

CTR (Click-Through-Rate) prediction has become an important link to the recommendation system. As deep learning techniques mature in computer vision and natural language processing, it is also beginning to be used in the area of recommendations. In this paper, we propose a new CTR prediction model based on xDeepFM with attentional mechanisms, attention mechanism can enhance the ability to represent characteristic information, and the importance of dynamic modeling features, the convergence ability of the model is optimized. When tested on the open data set, the model can effectively improve the accuracy of CTR prediction, stabilize the prediction deviation, suppress the influence of noise data, and can adapt to the CTR prediction task of any scene.