Research On Collaborative Filtering Model Based On Deep Learning In Recommendation System

Today,with the rapid development of the Internet era,the data has shown an exponential growth rate.In 2018,the number of Internet users worldwide has reached 3.6 billion,which will exceed 50%of the global population.The share of e-commerce is also growing,and the total value of Amazon's merchandise transactions has increased from 20%in 2013 to 28%in 2017.At the same time,the proportion of e-commerce recommendations is also rising,from 2%in 2015 to 6%in 201 7.How to make recommendationsmore satisfying the needs of consumers is the subject of our research.Nowadays,how to use social network data to make recommendations has become a hot issue in the field of recommendation systems.At the same time,collaborative filtering algorithms have become the mainstream recommendation algorithm.It is widely used in industry,the model is versatile,does not require much expertise in the field of data,and the project implementation is simpler and the effect is good.Of course,collaborative filtering also has some inevitable problems,such as the headache of "cold start",when there is no new user data,it is not good to recommend items for new users.This article takes the MovieLens movie dataset as the original dataset.Firstly,it introduces the collaborative filtering model,the implicit semantic model and the improved model.In order to improve the accuracy of the model prediction and the accuracy of the recommendation,the text is respectively used on the basis of the collaborative filtering model.Convolutional networks and long-and short-term memory neural network training models.The main contents of the paper are as follows:1.Introduce the collaborative filtering,implicit semantic model and the theoretical method and parameter derivation process of the improved model;2.Introduce the theoretical methods and parameter derivation of convolutional neural network and long-term and short-term memory neural network in deep learning respectively.3.Based on the MovieLens dataset,the recommended accuracy rate calculated by the improved collaborative filtering model is 25.2163%,and the recall rate is 12.1063%.The recommendation accuracy calculated by the implicit semantic model is 20.4613%,and the recall rate is 9.8263%.It can be clearly seen that the performance of the improved collaborative filtering model is better than that of the implicit semantic model;On the basis of the collaborative filtering model,the model is trained by text convolutional neural network and long-term and short-term memory neural network respectively,and the mean square error is used as the loss function to optimize the network.5.Calculated,text convolutional neural network of the training set loss error is 1.0202 and the test set loss error is 0.9905.The training set loss error of the long-term and short-term memory neural network is 0.8493,and the test set loss error is 0.8388.Whether it is to look at the training set loss error or the test set loss error,the loss error of the long-and short-term memory neural network is smaller,indicating that the long-and short-term memory neural network performs better than the text convolutional neural network.