Research On Spatial-temporal Clustering Algorithm Based On The Extraction Of User Trajectory Of Smartphone

With the popularity of smartphones, the use of mobile device to get the location of the sensor is more and more convenient, this paper, through the improvement of traditional clustering algorithm, the extraction of stop point. With Myeclipse as the main development environment, combined with Tomcat server and Java, the WebGIS based stop point extraction system is implemented, and the clustering results are statistically analyzed. The research of stop point has important academic significance and application value in the field of personalized friend recommendation, travel destination prediction, commercial advertising push and so on. In this paper, we focus on the spatial-temporal clustering algorithm based on the user trajectory point extraction and the system implementation of WebGIS.Data sources of this paper is the use of the phone APP obtain the same track four different Android phone users to track data. The actual stop point of the spatial-temporal travel by users of statistics, as a result of cluster analysis and evaluation of comparative data. Then, use a Java program to preprocess raw location data, including data cleansing, data centers and standardized three aspects to get latitude and longitude coordinates of the user trace points, time stamp and other information. Finally, the preprocessed data into the database according to a certain format to form the experimental data clustering analysis.Clustering algorithm based on Java language, were used based on the hierarchical of ST-BIRCH algorithm based on density ST-DBSCAN and ST-OPTICS algorithm and based on grid and density ST-GRID four spatio-temporal clustering methods were stay point extraction. According to the user's track of latitude and longitude range, the length of residence time, to stay the same point algorithm parameters set a specific time interval, to solve the problem of setting of algorithm parameters, extraction spatio-temporal stop point.In the evaluation of the comparison of different algorithms, by contrast with the time and space between the actual stay points from clustering algorithm to extract the correct point, deletions and point error point of these three aspects, and by comparing the different equipment, different algorithms, different types of targeting to get temporal residence temporal differences in points of difference algorithms to extract the results were analyzed and evaluated. General results of the spatio-temporal travel point extraction experiment: the same track at a different end of the phone different positioning strategy, staying point extraction accuracy significantly different; The Base Station Positioning error is one of the main factors that affect spatio-temporal travel point extraction accuracy, and GPS signals vulnerable shelter affected, prone positioning drift error, but remain little error on the extraction point; in terms of clustering algorithm based on density of ST-DBSCAN algorithm to track point density divided neighborhood, better able to noise removal, extraction better, but also because of other interpretative clustering algorithm, but the initial positioning error of The Base Station Positioning, is still difficult to eliminate; according to experiments, ST-OPTICS algorithm proposed a method to improve the accuracy of travel point extraction.The development of WebGIS system based on the API JavaScript application program interface of Baidu map is developed, which is combined with the clustering algorithm of Java language and the Oracle database, and the system is designed and implemented. The system uses three-division structure, the consistency of the system to meet the different levels of data transmission between the uniform interface. System to achieve the basic operation of the map, the user's actual stay information query, user trajectory query, clustering algorithm results show four major modules. Among them, the basic operation of the module that contains the map zoom, roaming, zoom, measure distance and other basic functions; The user stays information query include the user's actual stop point and the stop point buffer query; The user trajectory query module includes the user's trajectory point space information and attribute information query, the use of heat map to display the user's trajectory directly; Clustering analysis module is by different mobile phones, algorithm and parameter selection, combining with background Java language implementation of the clustering algorithm, the clustering results are returned to the browser, Clustering point query and stay point extraction, reverse address resolution and stay point of streetscape function, and through with the user residence information query module based on comparison of clustering algorithm comparison and evaluation.In this paper, the experimental results show that the main reason for the impact of the user's trajectory stay point extraction algorithm is that the base station location error is large. Based on the improved ST-OPTICS algorithm, for deciding on the minimization of the domain knowledge of parameters, determine the clustering threshold ST-DBSCAN algorithm of spatio-temporal stay point extraction, of traditional clustering methods have certain improvement for in spatio-temporal stay point extraction. However, the algorithm needs to be optimized in the time complexity and execution efficiency, and the integration of the WebGIS stay point extraction system is still need to be further studied.