Service Recommendation Method And Application System Based On Deep Learning

The development of the Internet provides people with more convenient services and more choices than before.While people enjoy the convenience and efficiency of online services,they often need to spend more time looking for services that are of real interest.The emergence of a personalized service recommendation system can help reduce the impact of information overload on the efficiency of service matching,and allow users to reduce the time it takes to find services that they are really interested in.The importance is self-evident.This article takes personalized service recommendation based on deep learning models as the main research content.The main work of the thesis can be divided into the following aspects:(1)Review and thinking of classic recommendation models: The recommendation system can be divided into pre-deep learning era and deep learning era.In the second chapter,the paper first reviewed the classic recommendation model in the pre-deep learning era-collaborative filtering family,analyzed the advantages and disadvantages of these models and applicable scenarios,and then introduced the later proposed machine learning model-factor decomposition machine(FM)How to solve these problems faced by the model of collaborative filtering family.Subsequently,in the third chapter,the paper studied the very classic and effective neural collaborative filtering(NCF)model,and used Tensor Flow to reproduce the NCF model and conducted experiments on the yelp dataset.(2)Research and experiment of recommendation models based on Convolutional Neural Network(CNN): At present,academia and industry sections have done a lot of research on recommendation systems based on deep learning,and many models have been proposed.In this paper,first,the CNNRec model is realized by extracting the features in the text comments using a convolutional neural network module(CNN).Experiments show that the use of CNN module to extract text comment features can effectively improve the model's effect on score prediction targets.On the basis of this research,the TS-CNNRec recommendation model is proposed in this paper to address the problem that the review text cannot be obtained in advance in the actual recommendation scenario.The model draws on the idea of the studentteacher network and uses the teacher network to learn the real comment text,and then guides the student network to learn effective user features and service features.The experimental results on the yelp data set show that the TS-CNNRec model's score prediction error when the comment is not known is significantly improved compared to the NCF model.(3)Finally,based on the proposed TS-CNNRec model,using the flask framework and other related technologies,a simple merchant service recommendation prototype system is implemented.