Research And Implementation Of Recommendation System Based On Deep Learning

In recent years,with the rapid development of big data and artificial intelligence,deep learning has been applied in all aspects of people's lives as an important tool for feature extraction.The recommendation system has become the main module in all mainstream applications due to its powerful data processing and information retrieval capabilities.This thesis focuses on the related techniques of deep learning in recommendation algorithms.In the hot challenge of the recommendation system,user click-through rate problem,the deep factorization machine is the best model.We introduce and analyze it,and propose the residual interaction model.At the same time,the design and implementation of the relevant recommendation system was carried out in the financial robot.The innovation and contribution of this dissertation are mainly divided into the following two parts:1.We analyze the recommendation algorithm based on deep learning.In the solution to explore the user click-through rate problem,the user and the item are the two subjects in the recommendation system,and the degree of association represents the user's biases on this item.However,the association information between users and items is often ignored.Therefore,we propose a "user-item interaction feature representation" based on the attention mechanism.The original feature is reconstructed by method of interactive representation and used for model training.In order to further extract more effective high-dimensional interaction information between users and items,this thesis deepens the fully connected network by using the idea of residual network.Compared with the ordinary network deepening,the residual network can extract higher-dimensional feature information on the basis of ensuring the robustness of the model.The residual interaction model is trained by the joint training of low and high-dimensional features.2.In financial robots,users often obtain target information through active consultation.In order to provide users with personalized information services,this thesis completed a recommendation system based on deep learning.The system is mainly divided into two modules,a data processing module and a recommended algorithm module.In the data processing module,we mainly introduces how to perform data crawling,data preprocessing,and data updating.In the recommendation algorithm module,it is mainly divided into three parts,a deep model recommendation,a collaborative filtering recommendation,and a content-based recommendation.Through the introduction of three recommended algorithms and application in the financial robot system,a recommendation system based on deep learning is finished.The experimental results show that the residual interaction model proposed in this paper performs best in the public data set;in addition,the related application of the deep model in the financial robot is realized.It is proved that the recommendation algorithm model based on deep learning is feasible.