Research On The Individulized Travel Routes Recommandation Technology Based On The Change Of User's Interests Features

Since the concept of Smart Planet is proposed, many countries started to invest in related research and development. Together with latest advance in information technology, tourism industry is attempting to accomplish various "smart" industrial transformations. Among them, travelling route personalisation becomes one of the most popular topics. With respect to the current research and application on travelling route personalisation, improvements are yet to be made. For example, the majority of related algorithms are based on mechanical recommendation methods such as traditional collaborative filtering, associative rules and n department chart. In terms of the practicality of recommendation results and closeness to individual user interest, they are not satisfying enough. While there exists some systems that involve online interaction with users and leverage user interest, they are not capable of learning and capturing the changing nature of user interest.In this thesis, user interest features for clues, personalized of recommendation and quality of result for goal. We study travel route to recommendation technology based on the change of user interest characteristic. The main research work summarized in the following three points:1) This thesis recommend a model based on the changes of user interest characteristic. we retrieved data on tourist attractions and travelling route using distributed web crawler to model user interest. Then, we extracted domain features on tourist attractions, travelling route and user behaviour and established feature model for user interest. Based on this, we calculated user-user similarity and user-attraction similarity.2) This thesis put forward collaborative filtering algorithms based on travel route popularity and user interest characteristic changes and vertex constraint recommendation algorithm based on user interest. We incorporated personalised user interest factor into the traditional collaborative filtering algorithm and introduced current popularity of travelling route as well as the weights of user interest feature changes with time sequence in the user-user similarity. Regarding to the problems in collaborative filtering, data sparsity problem and the problem of feature matching converging to zero, we based on Hongliang Lv's modification on PageRank algorithm. By adding nodes constraint set on user interest features, we proposed collaborative recommendation algorithm that takes into account user interest, tourist attractions and travelling routes.3) This thesis proposed a judgment calculation of data sparseness and personalized adjustment of recommendation algorithm parameters. We proposed a way to measure data sparsity and the level that feature matching converges to zero. We also set up feature base on users' changing interest and combined dynamic tuning on the recommendation parameters according to the changes of user interest features. As a result, better learning ability and more personalised recommendation become possible in our recommendation system.Experiment result showsed the introduction of travelling route popularity and user interest variation feature in collaborative filtering process improves both the recommendation efficiency and quality. Collaborative recommendation based on user interest, tourist attraction and travelling route is quite advantageous in terms of recommendation item number and quality under the circumstance of data sparsity and feature matching converging to zero. And after adding nodes constraint set on user interest features, recommendation efficiency improves by a large margin. By dynamically tuning the parameters and threshold of recommendation algorithm, recommendation result captures more on users' preference and personalised recommendation becomes realistic.