Improved Wide&Deep Recommendation Algorithm for User Behavioural Characteristics

In today's era of rapid data and information growth, people are exposed to more and more complex information. Recommendation system solves the problem of how to accurately deliver information to users, and also helps companies better understand user needs. According to the user's market demand, constantly improve the product experience. By analysing users' information in different areas, we can recommend products that meet their needs and interests in a multi-dimensional way. By analysing the user's behaviour, personal interests and other relevant information, we can improve the recommendation algorithm to push the items that the user is interested in, thereby improving the user experience and increasing the user's time of use.In recent years, many models based on Wide&Deep recommendation algorithms have been proposed.The Wide&Deep model can solve many problems of traditional algorithmic models, but at the same time there are also problems such as large data requirements and complex feature engineering. Based on this, a new Deep&iFFM model is proposed, which cross-fertilises each set of features to achieve the selection of different boosting methods for different attributes. The accuracy can be further improved by the collected dataset of user articles.