Research And Application Of Location-based Service Oriented Track Mining And Recommendation Technology

With the popularity of GPS and other positioning devices,outdoor track mining based on GPS has been studied and applied more and more.For example,knowledge such as user interest location can be mined from GPS track and applied to location-based recommendation system.But the GPS based location service technology can not meet the needs of indoor location service system.Indoor positioning requires technologies different from GPS,and indoor location information and trajectory routes are more complex than outdoor ones.Therefore,it is necessary to study technologies related to indoor location service,but there is little work in this field at present,and there are still many challenges to be faced.Aiming at the requirements of the development of indoor and outdoor integrated location service information system in a park,this paper studies the key technologies of indoor trajectory mining and location recommendation,and verifies the feasibility of relevant algorithms through the implementation of a prototype system.The main work completed in this paper is as follows.(1)Aiming at the problem of indoor trajectory analysis and mining of location service,this paper proposes a trajectory mining algorithm based on indoor pedestrian trajectory data.The original trajectory data are processed by the algorithm.Firstly,the user trajectory points are matched with the indoor location.Then,the location name of the user is obtained according to the indoor map data and the matched track points,that is,the track data with additional semantics.Finally,the interest area,aggregation area and key nodes in location semantic trajectory are mined.(2)Aiming at the accuracy of location recommendation of location service,this paper proposes a travel recommendation model that integrates self-attention mechanism and Long Short-Term Memory(LSTM)network structure.Modeling using the user long-term location data and short-term position,the user's attractions sequence data processing input data into a feature vector as a model,through the short-and long-term memory network structure and the attention mechanism model respectively to study the user long-term and short-term interests,preference,weather,and considering the user comments distance and other factors,will merge the attention mechanism and network structure of tourism both short-term and long-term memory recommended models to predict long-term favour attractions sequences and short-term fusion,reorganize,finally get the tourist route conforms to user preferences.Experimental results show that the proposed method can effectively improve the accuracy of recommendation.(3)For trajectory mining,and the proposed location of the application of recommendation algorithm,developed a certain park indoor location service system,the system is studied and analyzed the characteristics of indoor track data,the basis of design the interior trajectory data processing process framework,and implemented from the interior trajectory sampling data to the indoor location information related to extract the location of the abstract semantic trajectory,and then through the proposed path mining algorithm,semantic trajectory in the interest area,the position of the mining community,key points,such as knowledge visualization,position to recommend,the position such as path planning,resource allocation related services provided inspired characteristics related algorithm design.Through the application in the indoor location service system of the park,the prediction and decision performance of the whole system has been improved,and the expected effect has been achieved.