Research And Application Of Personalized News Recommendation Algorithm Based On Deep Learning

With the development of computer technology and the continuous popularization of the Internet,it meets the needs of users for information in the information age,but there is also the problem of information overload,and the personalized recommendation system is an effective way to solve this problem.In this paper,the recommendation algorithms based on improved NCF and LightGCN are studied,which integrate NCF with attention mechanism and LightGCN with contrastive learning techniques,respectively.This allows for more accurate capture of what users are interested in and improves the performance of the algorithm.NCF recommendation algorithm is widely used in recommendation system,it further applies deep learning to recommendation system,is a general framework,can express and promote matrix decomposition and use multilayer perceptron to learn user-project interaction function,greatly improve the performance of recommendation.In NCF,however,the effect of all features is considered the same,which dissolves a lot of valuable information.In this research work,it is proposed to integrate the attention mechanism into NCF.Specifically,it assumes that different cross-features have different degrees of influence on the results,considers the different degrees of influence of different features on the model,adjusts the weight of features,and optimizes the loss.The experimental results show that under the same experimental setting,the evaluation index compared with NCF is improved when the experiment is carried out on the same data set.LightGCN recommendation algorithm is a hot research topic in the recommendation system,it applies deep learning to the recommendation system,and simplifies the graph convolutional neural network applied in the recommendation system,which greatly improves the recommendation performance.However,the loss function in LightGCN is relatively single,and only BPR loss and L2 regularization loss are used,which fails to make full use of the dataset.In this research work,it is proposed to integrate comparative learning into the LightGCN model.Specifically,a boundary is set to filter information with low similarity,and hyperparameters are used to control the relative weight between positive and negative samples,maximize the similarity between positive sample pairs,and minimize the similarity between negative sample pairs that have been filtered,which makes the experiment better.The experimental results show that the evaluation index compared with LightGCN is improved when experiments are carried out on the same dataset under the same experimental settings.This article implements a small news recommendation system to improve the user experience,and it can also solve the problem that users cannot find the news they are looking for.The system adopts the combination of VUE and Python,and the VUE framework is mainly responsible for displaying pages.Python is mainly designed using the Django framework,which is mainly responsible for interacting with the database and integrating the two proposed recommendation algorithms into it to generate recommendation sequences.