Research On Personalized Medical Recommendation Method Based On Convolutional Neural Network

With the rapid development of the era of big data,data on the Internet has exploded.In the face of complex information,users cannot quickly select useful information for themselves.Recommendation system can actively recommend the content that users are interested in based on their interest preferences.In the past decades,it has been fully developed and widely used in shopping,tourism,food and other fields.At the same time,with the rapid development of deep learning technology,combining it with the traditional recommendation algorithm also creates new vitality for the development of recommendation system.In recent years,with the development of medical informatization on the Internet,a large amount of medical data has been accumulated on the Internet,resulting in a large number of medical data that cannot be effectively used,and personalized medical recommendation services are relatively few.Therefore,it is of great significance to study personalized medical recommendation.In order to improve the performance of medical recommendation,solve the mismatch between rating and evaluation in rating prediction and static coding of review text,this paper studies the deeply personalized medical recommendation based on the fusion of rating features and sentiment analysis of review text.The main work is as follows:1)Proposing a personalized recommendation method that integrates the characteristics of revised rating and review text.This paper presents the method of integrating rating and review text features to solve the problem of mismatch between users reviews and rating of doctors.Such as,users review of doctors is not good,but the rating is relatively high.This method combinates the original rating and review emotional analysis values to get the modified rating feature,which is selected as the final rating feature,and,the review text is represented by word2 vec embedding.Finally,the modified rating and review text are input into the model.The experimental results in the medical dataset show that this method make effect to alleviate the mismatch between rating and review in the medical field.2)Proposing a deep personalized medical recommendation based on the fusion of emotional characteristics of rating and comment.This paper uses BERT pre-training to solve the inaccuracy of review representation,and splices the depth features of medical text by combining deep learning method.For medical review text,it first used BERT pre-training to vector representation.Then,the review text of vectorization representation splices the modified rating feature as input data which is inputted into convolution mind.Finally,a factor decomposer is used to interact with the features and the root mean square error is used to measure the model performance.Through four groups of comparison experiments,it is verified that:(1)The effect of BERT is better than word2vec.(2)The method of using deep learning is better than the traditional method.(3)The performance of double-layer network is better than that of singlelayer network.(4)Through ablation experiments,it is proved that the effect of embedding comment text by using the modified scoring feature and Bert is better than that by using one of them alone.