High-precision spatio-temporal trajectory recovery method based on cell phone signaling data

A high-precision spatio-temporal trajectory recovery method based on cell phone signaling data, applied to a computer, comprising:

• using a density-based spatial clustering of application with noise (DBSCAN) to remove discrete point, using a rule-based method to remove ping-pong data which is data generated during Ping-pong handover, and using a grid clustering algorithm that fuses spatio-temporal features to identify a staying point of the user, the method further comprising:
• pre-processing original signaling data by using a data pre-processing method based on the DBSCAN;
• using the grid clustering algorithm that fuses the spatio-temporal features to identify the staying point of the user;
• and
• after identifying the staying point of the user, removing an identified staying point, and further reconstructing an user trajectory;
• wherein the pre-processing original signaling data collected by using the data pre-processing method based on the DBSCAN comprises:
• selecting a complete travel record of a user, and marking missing data as 0 point;
• in response to that there is the 0 point, determining whether a previous timestamp record of the 0 point and a next timestamp record of the 0 point are equal, and in response to that they are equal, deleting the 0 point, and in response to that they are not equal, replacing the 0 point with an average value of the previous timestamp record and the next timestamp record;
• after deleting the 0 point, processing drift data by using the DBSCAN, and processing the ping-pong data by using the rule-based method, and removing the drift data by a method of setting a time window;
• wherein the drift data is anomalous positioning data due to a device's inability to connect to a nearest base station, and the drift data is usually not in the same or neighboring grid with a previous trajectory point, which is manifested by the inability of the previous trajectory point to continue backward in a grid;
• in response to that a first trajectory point that cannot continue to be linked backward, and there is a second trajectory point that continue to be linked backward within the time window of a downward direction of the first trajectory point;
• removing all trajectory points between the first trajectory point and the second trajectory point, wherein all trajectory points between the first trajectory point and the second trajectory point are the drift data, otherwise continuing downward to search by the time window with the previous trajectory point of the first trajectory point a reference;
• for the ping-pong data, gridding a studied cell, mapping remaining trajectory points after drift data processing into a grid, setting a time threshold according to characteristics of the ping-pong data that jumps rapidly between neighboring base stations, and in response to that the trajectory points in the grid jumps between two base stations and a time interval for the jump is less than the time threshold, determining the trajectory point to be the ping-pong data and removing the ping-pong data;
• wherein the using the grid clustering algorithm that fuses the spatio-temporal features to identify the staying point of the user comprises:
• gridding a studied traffic cell;
• mapping cell phone signaling trajectory points according to a time sequence into a studied traffic cell grid;
• calculating user's continuous residence time in each grid;
• determining whether time difference between time the user leaves the grid and time the user arrives at the grid is greater than the time threshold;
• and
• in response to that the time difference is greater than the time threshold, determining the data to be residence data, and otherwise determining the data to be travel data;
• wherein for the residence data, a concept of points of interest (POI) is introduced, the POI is captured by Baidu to obtain information comprising name, type, latitude and longitude to further analyze purpose why the user resides, for subsequent study of the user's travel behavior;
• wherein the grid clustering algorithm that fuses the spatio-temporal features comprises:
• gridding the studied traffic cell, and mapping the user trajectory point into the studied traffic cell grid according to the time sequence;
• defining the pre-processed signaling dataset as userData, a center latitude and longitude of the gird being gridSizeLng, gridSizeLat respectively, and setting the time threshold to be T;
• initializing all data groups, and sequentially determining whether the pre-processed data are all in the grid, in response to that the pre-processed data are all in the grid, adding the trajectory point to a grid cluster group, in response to that the pre-processed data are not all in the grid, adding the trajectory point to a grid cluster all group to distinguish between the two data, and then continuing to determine a next trajectory point until all the data are all determined;
• and
• in response to that the data within the grid cluster is within the grid cluster all group and time difference between time the trajectory point leaves the grid and time the first trajectory point rives at the grid is greater than the time threshold T, determining the data to be stay data and adding the data to a stayCluster group, to obtain the resident's staying point;
• wherein in an operation of the after identifying the staying point of the user, removing the identified staying point, and further reconstructing an user trajectory, on a basis of the identification of the user's staying point, analyzing the user's behavior by combining road network data collected by Maps, and analyzing the user staying area and staying purpose according to the user's POI data, to facilitate the analysis of the user's subsequent travel behavior;
• dividing the user trajectory into multiple sub-trajectories at equal distances;
• and
• determining a travel mode of the user by clustering the user's travel data within each sub-trajectory.