Research On Recommendation Algorithm Based On Recurrent Neural Network

With the rapid development of technologies such as cloud computing and big data,more and more applications in the Internet have caused the data scale to explode.As an effective method to solve the problem of information overload,the recommendation system has been widely used in many fields.However,the traditional recommendation algorithm mainly uses shallow models or artificially extracts features to learn features,resulting in the algorithm not being able to obtain deeper features of users and projec ts.Moreover,the traditional recommendation algorithm mostly considers that the user's attribute is fixed,ignoring the influence of the dynamic change of the user's interests in the sequence data of the user behavior on the recommendation result.In recent years,deep learning has made great achievements in the fields of computer vision and speech recognition,and it has also brought new opportunities to the recommendation system.Deep learning can learn more essential features of users and items from massive user behavior data through its deep nonlinear network structure.The recurrent neural network can model the user's dynamic changes of interest by modeling the sequence data of the user behavior,and improve the recommendation accuracy of the next mome nt.Therefore,this paper focuses on the research of recommendation algorithms around the recurrent neural network.Firstly,it is considered that the current recommendation algorithm based on the recurrent neural network only pays attention to the sequence information of the user behavior sequence,ignoring the user's preference for the item,and proposes a recurrent neural network recommendation algorithm based on the rating preference.In addition,a recurrent neural network recommendation algorithm based on deep semantic structured model is proposed by combining the recurrent neural network and the deep semantic structured model.The specific research contents are as follows:1.A recurrent neural network recommendation algorithm based on rating preference is proposed.The algorithm models the user behavior sequence data through the recurrent neural network,and then takes the user's rating of the item as a preference for the corresponding item in the sequence,and then combines it into the prediction of the next moment item.By testing on the real data set,the results show that the method recommended to the user at the next moment is not only the item that the user will click,but also the item that the user is interested in.2.A recurrent neural network recommendation algorithm based on deep semantic structured model is proposed.The algorithm combines the recurrent neural network and the deep semantic structured model.On the one hand,the user behavior data is modeled by the recurrent neural network to learn the implicit representation of the user's current interest.On the other hand,the model parameters are reduced by the deep semantic structured model,which speeds up.The convergence of the model.At the same time,in order to let the algorithm learn more effective user feature representation,a recurrent neural network recommendation algorithm based on deep semantic structured model with user features is proposed.Through multiple sets of experiments,the algorithm improves the recommended hit rate while converging faster.