A Tourism Service Recommender System Based On Deep Learning

In recent years,with the popularization of the Internet and the continuous development of information technology,people's demand for cultural tourism has become more and more diversified.Effective tourism service recommendation is of great significance for providing efficient and high-quality personalized tourism service recommendations.However,while the traditional tourism service recommendation improves the quality of tourism services,the hidden problems also restrict the further improvement of service quality.On the one hand,the traditional recommendation method relies on the design of shallow features and cannot fully and effectively learn the deep features of users and items.On the other hand,it is impossible to fuse well multi-source heterogeneous data,and the data sparsity problem cannot be well alleviated.The above two issues ultimately affected the prediction accuracy of travel service recommendations.An effective method is to use deep learning techniques to improve the accuracy of travel service recommendations.Based on the above ideas,the main work of this paper is as follows:1.For most of the recommendation systems that use textual information without considering the semantic similarity and other factors that can not obtain important comprehensive features,this article uses the word embedding method to process comments and other textual information in the data preprocessing stage.This method can use low-dimensional,Dense,real-valued vectors represent each word,not only can be a rich text feature expression,but also integrate the semantics into the model to achieve the use of text semantic information.2.In this paper,two convolution networks(CNN)and a deep neural network(DNN)are designed and constructed by studying the difficulties in extracting information features from shallow learning models in common recommendation algorithms,difficult to learn deep knowledge,and unable to automatically acquire features.Based on the advantages of the CNN network in extracting text features,the CNN network is designed to process user review information and travel service project review information.The network is designed as a four-layer structure,namely input layer,convolution layer,pooling layers and full connection layers.Because the deep neural network DNN can fuse multi-source heterogeneous data well,the deep neural network(DNN)is designed to deal with the basic information of users and tourism service projects.The network is designed as a three-layer structure,which is the input layer and two fully connected layers.3.At the end,a common interaction layer is constructed to connect the three networks together,and the layer design uses a factorization machine.The main purpose is to improve the accuracy of the prediction by learning the interaction between the features extracted from the three networks,to a certain extent,to alleviate the impact of data sparsity,and to further improve the establishment of the model.4.By studying the above problems and combining the methods mentioned above,a set of tourism service recommendation models based on deep learning was constructed and implemented.By comparing it with common traditional recommendation methods,it can be found from the experimental results that the proposed model method is obviously better than the traditional recommendation method,which verifies the superiority of the model.Further experiments are compared with different depth model variants.The experimental results show that the model combination also has better performance than the depth model of other variants,which verifies the effectiveness of the model combination.This laid a good foundation for the promotion and use of tourism service recommendation system.