Research On Tourism Information Recommendation Based On Spatio - Temporal Data

The rapid development of Mobile Internet technology and location information collectiontechnology makes the network become the main channel of people to access and shareinformation. How to search information from a large amount of data to meet users’ demands?It has become the hot topic in the field of study. Especially in the tourism industry, people shareexperiences on social networking sites by sharing photos, which generates a lot of time andspace data. The spatio-temporal data mainly includes geographical location (longitude andlatitude), information such as shooting time and the description of photos description. It isextremely important for analysis mobile users travel behavior, mining user interest preferencesand providing users with attractions.At present, the recommendation of information for travel is concentrating on two parts:personalized travel recommendations, and classic travel recommendations. Among them, classicrecommendation is only analysis popular attractions in tourist destination. Personalizedrecommendation is based on user history travel path. Attractions are recommended to the user inaccordance with the preferences of attractions as well as the optimal tourist route planning. Butthe current study has the following problems: The user similarity calculation is not accurate andhas high computational complexity; Not considering the mutual influence of user’s preferenceand popular attractions; Without considering the influence of the user’s current contextinformation to recommendation results; Not considering the weather conditions of the target cityattractions to the user choice. Aiming at the above problems, we make the following work:(1) Improved the user similarity algorithm. This paper puts forward the concept of coreusers. According to the label information of photos, we increase the semanticinformation for scenic spots and mining users’ favorite attractions types. Similaritycalculation is on the basis of core users’ matrix and attractions’ types, we reduces thecomputational complexity and improve the efficiency and accuracy of the algorithm.(2) Put forward a recommendation algorithm based on users’ interests and scenic popularity,which is called CIAP. CIAP defines user similarity function and attractions populardegree function and gets the optimal results of recommendation by determiningsimilarity of user’s attractions weight and attractions popularity weight.(3) A combination of context awareness recommendation algorithm. In this paper,according to user’s current context awareness of time and weather information, we filterthe attractions of recommendation and recommend suitable attractions for the user.We carry out experiment by using the data obtained from Flickr open interface. We evaluatethe recommended effect of CIAP algorithm in different weight. The result of experiment showsthat the CIAP algorithm has better performance of comprehensive recommendations. We alsocompare our similarity algorithm with others, the result prove that our recommendationalgorithm has a better effect.