The Algorithm On Personalized Travel Attraction Recommendation Based On Transfer Learning

In recent years,with the rapid development and popularization of Internet technology,users have generated massive amounts of data information while acquiring data from the network.In the tourism industry,as the data on the network grows spurt,on the one hand,users can obtain more abundant information on the spot.On the other hand,when users face the majority of tourist information,the user needs to spend a lot of time retrieving really valuable information.The problem of “information overload” has become more and more serious,and the traditional tourist attraction recommendation system has been unable to meet the needs of users.Therefore,a tourist attraction recommendation system that can meet the needs of users has been highly praised by users,namely,a personalized recommendation system for tourist attractions.However,the recommendation system generally has three problems of cold-start,data-sparsity and low-accuracy.Therefore,this paper introduces transfer learning and deep learning algorithms for these problems,and conducts personalized tourist attractions recommendation based on domain adaptation and personalized tourist attractions recommendation based on deep transfer.The main research contents of the thesis are as follows:(1)The algorithm on personalized tourist attractions recommendation based on domain adaptation is studied.Aiming at the common problems in the recommendation system,personalized recommendation algorithm for tourist attractions based on domain adaptation is proposed.In the field of personalized tourist attractions recommendation,the target domain data in the target task is mostly unlabeled data,and the model cannot be trained,but the source domain data related to the target domain data is completely labeled,so consider that the source domain dataset is introduced to assist the target domain dataset for model training.Firstly,the experimental auxiliary dataset is prepared,namely,the source domain dataset,and crawl the labeled data related to the target domain dataset from the network as the source domain dataset.Secondly,extract the image features of the target domain dataset and the source domain dataset.The method of Bag-of-Visual-Words model(BoVW)is adopted for extracting features;then,because there is a difference in data distribution between the target domain dataset and the source domain dataset,the features cannot be directly trained after extracting features.The domain adaptation technique is used to reduce the distribution difference between the target domain dataset and the source domain dataset,and then the support vector machine(SVM)classifier is used to classify the interest of the user,and finally the relationship between target users and others users is analyzed,in order to achieve personalized recommendation of tourist attractions based on domain adaptation.(2)The algorithm on personalized tourist attractions recommendation based on deep transfer is studied.In order to obtain a personalized recommendation system for tourist attractions with better performance,a deep learning algorithm is introduced,and personalized tourist attractions recommendation based on deep transfer is proposed.Compared with the traditional machine learning algorithm,the deep learning algorithm has the advantage of extracting image features with high robustness and discriminativeness,and then more accurately obtaining the user's interest,thereby improving the performance of the personalized recommendation algorithm.In order to solve the cold-start and data-sparsity problems,the source domain data is used to assist the target domain data.Because of the distribution difference between the target domain data and the source domain data,an adaptive layer is added to the convolutional neural network.The adaptation layer obtains the distribution difference between the data,that is,the domain loss,and then minimizes the domain loss to realize the deep feature transfer,thereby implementing the training of the model to obtain the attractions of interest to the user,and finally the relationship between target users and others users is analyzed,in order to achieve personalized recommendation of tourist attractions based on deep transfer.(3)The recommendation performance of the proposed algorithm is verified by experiments.The experimental results show that personalized tourist attractions recommendation based on domain adaptation can solve the cold-start and datasparsity problems well,and the recommended performance is good.Moreover,personalized tourist attractions recommendation based on deep transfer not only solves the problem of data-sparsity,but also improves the recommendation performance obviously.