Research On Recommendation Methods For Combining Static Features And Dynamic Features

The rapid growth of information makes it difficult for people to quickly find the information they need from the vast amount of data.It is difficult to effectively solve the problem by the application represented by the portal and search engine.The recommendation system provides a feasible solution to the problem.At the same time as a means to help users quickly retrieve information,the recommendation system has also brought huge benefits to many companies.Therefore,it has attracted the attention of many scholars,and relevant research results have emerged in an endless stream.In-depth study of the recommended methods,to explore a recommendation algorithm with higher accuracy,so as to efficiently recommend appropriate information for users,has important theoretical value and practical value.The general flow of the recommendation method based on deep learning is given,which is divided into three stages: feature extraction,preference prediction and recommendation.The intrinsic characteristics of users and articles are analyzed.It is pointed out that in most cases,it is difficult to obtain the inherent attributes of users based on privacy considerations.Currently,the research of recommendation system is mainly based on the interaction data of users and items.The preferences and information about the user and the item itself are often an important part of this type of data.This recommendation method with preferences,users,and items as the core applies to a variety of scenarios,including movie recommendations,news recommendations,product recommendations,and more.Under this premise,the interaction and behavior data of users and items are analyzed and classified,and the interactive behavior data of users and items are analyzed from static and dynamic perspectives,and the static and dynamic characteristics of users and items are extracted.A specific extraction method was developed,and the two types of features were merged for the recommended requirements.Based on the user and item characteristics of the fusion,a user's preference prediction model for the item is designed,and the corresponding recommendation method is given according to the prediction model.Experiments were carried out on the proposed method.In order to facilitate the development of the experimental work,only a representative data set such as the movie type were used in the experimental session.The results show that the method has a better recommendation effect to some extent,and the verification is verified.At the same time,the influence of parameters such as feature dimension and sequence length on the recommendation effect is discussed.