Research And Implementation Of Text Classification And Recommendation System Based On The Deep Learning

Along with the advent of the Internet+ era,the more and more applications associated with the Internet,such as smarter transportation,the network shopping,share the rent and the recommendation system of information and knowledge,and so on,have brought about great changes to our lives.Particularly recommended system provides us with a personalized service,and reduced the originally spend a lot of the retrieval time for access to information.Now,however,the recommendation system accuracy neither is high enough,intelligent nor meets the demand of people.Therefore,building a higher accuracy of intelligent recommendation system is a problem to be solved.This thesis proposes a new reading oriented application based on the technology of deep learning intelligent recommendation system.Firstly,combined the technology of deep learning with a document theme probability model algorithm,this thesis proposes a hybrid model of text classification.The model using the LDA algorithm to calculate the probability distribution document theme,and then the topic distribution vector as input,use deep learning algorithms to classify training and prediction.The experimental results show that the classification forecast accuracy is higher than the Naive Bayes algorithm is about ten percent.Secondly,this thesis designs and implements a recommendation system for reading based on a classification algorithm.The system is built on a distributed computing platform Spark and distributed storage platform HDFS,including system recommendation module and system function module,greatly improve the scalability and performance of the system.In the end,the recommendation system for reading can provide users with accuracy of 3%above the collaborative filtering recommendation results,and the system has a better robustness and better expansibility.