JP2015049681A - Device, program and method for retrieving user taking similar moving route - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device retrieving a user taking a moving route similar to a moving route of a user carrying a mobile terminal by use of base station position information without depending on a positioning function of the mobile terminal.SOLUTION: A device has: means determining a short-time remaining place and a staying place on the basis of a probability distribution of a position based on base station position information in a first time window and a second time window in a communication history, and generating a staying/remaining time series data; means constructing an inter-short time remaining place and staying place-short time remaining place transition model associated with appearance probability of a transition pattern in each inter-short time remaining place transition pattern and each staying place-short time remaining place transition pattern; means interpolating the short-time remaining place lacking in a time series on the basis of the constructed transition model; and means retrieving a mobile terminal taking a similar moving route on the basis of a longest matching partial series length between time series data after interpolation processing and time series data after the interpolation processing related to another mobile terminal.

Description

  The present invention relates to a technique for estimating information related to stay and movement of a user who has a mobile terminal.

  In recent years, many mobile terminals typified by mobile phones have a positioning function such as GPS (Global Positioning System). Thereby, the user can acquire the current position information through positioning, and can receive various service information corresponding to the position by transmitting the positioning result to the server via the network.

  However, if the GPS function and its application are activated at all times or regularly in the mobile terminal in order to acquire the user's location information, not only will the battery of the mobile terminal be depleted, but also the amount of packets sent from the mobile terminal Will increase. Therefore, it becomes difficult to spread services based on user location information.

  Therefore, on the telecommunications carrier side, it is possible to estimate the position and time interval related to the stay of the user who owns the mobile terminal based on the position information of the base station under the control of the mobile terminal, not the position information acquired by the mobile terminal Is preferred. In this case, the mobile terminal does not need to activate the GPS function constantly or periodically.

  For example, Patent Document 1 discloses a technique for analyzing a moving line of a mobile body using a history of position information of a base station under the control of a mobile body such as a mobile terminal. According to this technique, the presence or absence of contact between moving bodies is evaluated based on the coincidence of staying places in two moving bodies and the overlap of staying time zones.

JP 2006-85257 A

  However, with the conventional technology described in Patent Document 1, it is difficult to detect whether one mobile terminal and another mobile terminal are accompanying each other or passing close to each other. For example, when a user who has a mobile terminal is moving by a moving means such as a train or a car, whether or not a user who has another mobile terminal is on the same moving means as the user in a predetermined time interval. Alternatively, it is difficult to estimate whether or not the moving means on which another user has moved is moving near the moving means on which the user has been boarded.

  In order to detect and estimate such traffic at the same location or close proximity, it is important to derive the travel routes of each user and compare them appropriately.

  Therefore, the present invention provides a user who takes a movement route similar to the movement route of the user who possesses the mobile terminal using the base station position information that can be acquired by the carrier equipment without relying on the positioning function of the mobile terminal. An object is to provide an apparatus, a program, and a method that can be searched.

According to the present invention, an apparatus for searching for a user who takes a movement route similar to the movement route of a user possessing a portable terminal,
Location information history storage means for storing a communication history including a plurality of communication records in which date and time related to communication and base station location information are associated with each portable terminal;
For each portable terminal, a “short-time staying place” is determined based on a probability distribution of positions based on a plurality of base station position information in the first time window in the communication history, and a time longer than this first time window A “stay place” is determined based on a probability distribution of positions based on a plurality of base station position information in a second time window having a width, and the determined “stay place” and / or “short stay place” is determined. Time series data generating means for generating time series data of staying residence arranged in time series;
“Short-term time” in which transition patterns between “short-time staying places” that occur at a predetermined frequency or more are extracted from the staying residence time-series data, and the transition pattern appearance probability is associated with each transition pattern between “short-term staying places” A transition model construction means for building a transition model between the residence areas for a short time,
The transition pattern between “stay place / short-time stay place” that occurs at a predetermined frequency or more from the stay stay time series data is extracted, and for each transition pattern between “stay place / short stay place”, the appearance probability of the transition pattern is calculated. A transition model construction means for constructing a transition model between the associated "stay place / short stay area",
Based on the built-in transition model between “short-time staying places” and the transition model between “staying place / short-time staying places”, the “short-term staying place” missing in time series Short-term staying ground interpolation means for interpolation;
Based on the longest matching partial sequence length between the stay and stay time series data after interpolation processing related to a predetermined mobile terminal and the stay and stay time series data after interpolation processing related to another mobile terminal, There is provided an apparatus having user search means for searching for a portable terminal that takes a movement route similar to the movement route.

The time-series data generating means in the apparatus of the present invention is:
Time window dividing means for dividing the communication history by each of the first time window and the second time window;
For each first time window, it is determined whether or not “staying” based on a probability distribution of positions based on a plurality of base station position information, and the probability of the position in the first time window determined as “staying” Based on the distribution, the “staying place” is determined, and the first having a different “staying place” before or after a predetermined time or more of the staying with respect to a predetermined first or less continuous time window determined as “staying”. If there are two time windows and a different “stay place” is separated from the “stay place” of the continuous first time window by a predetermined distance or more, the continuous first time window is “short”. First short-term residence determination means for determining "time residence";
Based on the probability distribution of the “staying place” in the first time window determined to be “short staying” for a plurality of mobile terminals, the “short staying place” is determined, and the base for each “short staying place” Base station information collecting means for generating a base station list in which station position information is associated;
When there is a base station included in the base station list for the communication history, the short-time stay location history in which the base station position information of this base station is associated with the “short-time stay location” corresponding to the base station list. Second short-term residence determination means for generating
For each second time window, it is determined whether or not it is “stay” based on the probability distribution of positions based on a plurality of base station position information, and the probability distribution of positions in the second time window determined as “stay” A stay determination means for determining a “stay place” based on
For each mobile device, based on the generated short staying place history, the stay staying time series data that generates the determined staying place and / or “short staying place” in time series is generated. It is also preferable to have series data generation means.

  Further, in the above-described embodiment, the base station information collection unit generates a base station list in which the base station position information of the base station and the connection probability with respect to the base station are associated with each “short staying place”. It is also preferable that the second short-term stay determination means selects only a base station having a connection probability equal to or higher than a predetermined value and generates a short-term stay location history. Further, the time window dividing means sets the average value, mode value, or N quantile (0 ≦ N ≦ 1) value of the short-duration time length in the predetermined area range as the time width of the first time window. It is also preferable to do.

  In addition, the short-time staying place interpolation means in the apparatus of the present invention, when there are a plurality of candidates of “short-time staying place” missing in time series, the appearance probability of the transition pattern including one candidate in between. It is also preferable to interpolate with the candidate where the likelihood based is highest.

  Further, this short staying place interpolation means, when the staying time series data adjacent "short staying places" or adjacent "stay place" and "short staying place" are separated by a predetermined time or more, It is also preferable to estimate that there is a missing “short staying place” between “short staying places” or between “staying place” and “short staying place”.

  Further, the user search means in the apparatus of the present invention calculates the similarity between two portable terminals by (longest matching partial sequence length) / (of the staying residence time series data after interpolation processing related to these two portable terminals) It is also preferable to search for a mobile terminal having a similarity equal to or greater than a predetermined threshold as the longer sequence length). Further, the user search means may investigate the longest matching partial sequence length for a time section between the time sections corresponding to the “stay places” separated from each other by a predetermined time or more among the specified search target time sections. preferable.

Moreover, according to the present invention, there is provided a communication equipment device as described above connected to a wide area wireless communication network,
Base station location information management means for storing a base station identifier and base station location information in association with each other;
From a base station to which the mobile terminal is connected under control, further includes a communication history collection means for collecting a communication record in which a date and time related to communication for each mobile terminal is associated with a base station identifier of the base station,
The location information history storage means provides a communication facility device for generating a communication history by further associating the base station location information corresponding to the base station identifier for each portable terminal with respect to the communication record.

According to the present invention, there is further provided a program for causing a computer installed in the apparatus to function so as to search for a user who takes a movement route similar to the movement route of the user possessing the mobile terminal,
Location information history storage means for storing a communication history including a plurality of communication records in which date and time related to communication and base station location information are associated with each portable terminal;
For each portable terminal, a “short-time staying place” is determined based on a probability distribution of positions based on a plurality of base station position information in the first time window in the communication history, and a time longer than this first time window A “stay place” is determined based on a probability distribution of positions based on a plurality of base station position information in a second time window having a width, and the determined “stay place” and / or “short stay place” is determined. Time series data generating means for generating time series data of staying residence arranged in time series;
“Short-term time” in which transition patterns between “short-time staying places” that occur at a predetermined frequency or more are extracted from the staying residence time-series data, and the transition pattern appearance probability is associated with each transition pattern between “short-term staying places” A transition model construction means for building a transition model between the residence areas for a short time,
The transition pattern between “stay place / short-time stay place” that occurs at a predetermined frequency or more from the stay stay time series data is extracted, and for each transition pattern between “stay place / short stay place”, the appearance probability of the transition pattern is calculated. A transition model construction means for constructing a transition model between the associated "stay place / short stay area",
Based on the built-in transition model between “short-time staying places” and the transition model between “staying place / short-time staying places”, the “short-term staying place” missing in time series Short-term staying ground interpolation means for interpolation;
Based on the longest matching partial sequence length between the stay and stay time series data after interpolation processing related to a predetermined mobile terminal and the stay and stay time series data after interpolation processing related to another mobile terminal, There is provided a program that causes a computer to function as user search means for searching for a portable terminal that takes a movement route similar to the movement route.

According to the present invention, there is further provided a method for searching for a user who takes a movement route similar to the movement route of the user who possesses the mobile terminal using the device,
The apparatus includes, for each mobile terminal, location information history accumulation means that accumulates a communication history including a plurality of communication records in which date and time related to communication and base station location information are associated with each other.
For each portable terminal, a “short-time staying place” is determined based on a probability distribution of positions based on a plurality of base station position information in the first time window in the communication history, and a time longer than this first time window A “stay place” is determined based on a probability distribution of positions based on a plurality of base station position information in a second time window having a width, and the determined “stay place” and / or “short stay place” is determined. A first step of generating time series data for staying and staying arranged in time series;
“Short-term time” in which transition patterns between “short-time staying places” that occur at a predetermined frequency or more are extracted from the staying residence time-series data, and the transition pattern appearance probability is associated with each transition pattern between “short-term staying places” A second step of constructing a transition model between “staying places”;
The transition pattern between “stay place / short-time stay place” that occurs at a predetermined frequency or more from the stay stay time series data is extracted, and for each transition pattern between “stay place / short stay place”, the appearance probability of the transition pattern is calculated. A third step of building a transition model between the associated “stay place / short stay place”;
Based on the built-in transition model between “short-time staying places” and the transition model between “staying place / short-time staying places”, the “short-term staying place” missing in time series A fourth step of interpolation;
Based on the longest matching partial sequence length between the stay and stay time series data after interpolation processing related to a predetermined mobile terminal and the stay and stay time series data after interpolation processing related to another mobile terminal, And a fifth step of searching for a mobile terminal that takes a movement route similar to the movement route.

  According to the apparatus, program, and method of the present invention, the base station position information that can be acquired by the telecommunications carrier equipment without relying on the positioning function of the mobile terminal, is similar to the movement path of the user who has the mobile terminal. It is possible to search for a user who takes a movement route.

It is the schematic which shows the stay and movement of a portable terminal in real space. It is a functional block diagram of the similar movement path | route user search apparatus by this invention. It is a table | surface of base station position information. It is a table | surface of communication history. It is a table | surface of the communication history which matched base station position information with the communication record. It is a table | surface which shows the start time and end time of a 1st time window. 7 is a table in which each row (communication record) in FIG. 5 is classified for each first time window in FIG. 6 for each portable terminal. It is a table | surface which shows the start time and end time of a 2nd time window. 9 is a table in which each row (communication record) in FIG. 5 is classified for each second time window in FIG. 8 for each portable terminal. It is a flowchart for determining whether it has unimodality. It is a table | surface which shows the 2nd time window determined as "stay", and corresponding representative point position information. 12 is a table showing representative point position information of a “stay place cluster” calculated based on representative point position information for each second time window shown in FIG. 11. It is a table | surface which shows the 1st time window determined as "staying", and corresponding representative point position information. FIG. 14 is a table showing the first time window determined to be “short stay” and the corresponding “stay place” representative point. FIG. 6 is a schematic diagram illustrating a state where information related to “stay” is collected from a plurality of portable terminals 2 and a table illustrating calculation results of stay place representative points in “short stay place”. It is a table | surface which shows the base station list | wrist of a short stay place. It is a table | surface which shows the result of a 2nd short time residence determination. FIG. 6 is a schematic diagram showing how the mobile terminals 00001 and 00002 progress while performing “stay” and “short stay”, and a table showing the stay stay time series data generated. It is a table | surface which shows the transition model between "short-time staying places". It is a table | surface which shows the transition model between "stay place and short stay place". It is a table | surface which shows the transition model between short-time stay places about short stay places AG. It is a table | surface which shows the stay place and short-term stay place transition model about stay places (alpha) and (beta) and short stay places AG. It is a table | surface for demonstrating the Example of a "short-time stay place" interpolation.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a schematic diagram showing the stay and movement of a mobile terminal in real space.

  According to FIG. 1, a mobile terminal (for example, a mobile phone or a smartphone) 2 possessed by a user is under the control of any base station at any position, and normally continues to communicate wirelessly with the base station. Can do. For this user, home and work are “stay places”. Further, stations A, B, C, and D, which are places other than home and work, serve as “short-time staying places” for getting on and off the train or by stopping the train.

  The telecommunications carrier equipment (FIG. 2) that integrates a large number of base stations 3 can always collect base station position information with a coarse spatial granularity and a constant time interval for each mobile terminal 2. Here, “the spatial granularity is coarse” means that the geographical distance between the position information is relatively long. Further, “the time interval is not constant” means that the acquisition time interval of the position information varies relatively depending on the communication timing.

  The base station 3 connected to the wide area wireless communication network (mobile phone network) acquires the date and time (for example, the date and time when communication was started) when communicating with the mobile terminal 2 located under the base station 3 as a communication history. The communication history is acquired when calling, sending / receiving mail, or browsing a Web page. Furthermore, it is acquired also at the time of communication between the application installed in the portable terminal 2 and the server, downloading or uploading of the application or content, and the like.

  FIG. 2 is a functional configuration diagram of the similar movement route user search device according to the present invention.

  The apparatus 1 according to the present invention can search for a user who takes a movement route similar to the movement route of the user who possesses the mobile terminal 2 using a communication history stored in advance. For example, when a user who has a mobile terminal is moving by a moving means such as a train or a car, whether or not the user who has another mobile terminal is on the same moving means as the user in a predetermined time interval. And a user who takes a similar movement route can be searched. Alternatively, it is also possible to search for a user by estimating whether or not the moving means on which another user has boarded is moving near the moving means on which the user has boarded.

  By estimating users who take similar movement routes in this way, for example, online information such as advertisements and coupons according to the movement route is distributed to a user group possessing a portable terminal and having a similar movement route. It becomes possible. The device 1 may be a communication equipment device that is installed in a wide area wireless communication network (mobile phone network) and collects communication histories from the base station 3 as appropriate.

  According to FIG. 2, the similar travel route user search device 1 includes a time window division unit 121, a stay determination unit 122, a first short time residence determination unit 123, a base station information collection unit 124, 2 short stay dwell determination unit 125, stay dwell time series data generation unit 126, short stay dwelling place transition model construction unit 127, and stay place / short stay dwelling place transition model construction unit 128. These functional configuration units function as the time-series data generation unit 12. The apparatus 1 also includes a short stay location interpolation unit 131 and a user search unit 132.

  The functional configuration unit described above is realized by executing a program that causes a computer mounted on the apparatus to function. In addition, according to FIG. 2, the flow of processing using each functional component is understood as a method for searching for users who take a similar movement route.

  Further, when the device 1 is a communication equipment device installed in a wide area wireless communication network (mobile phone network), a communication interface unit 10 connected to the wide area network and a communication history collection unit in order to acquire a communication history 111, a base station location information management unit 112, and a location information history storage unit 113. Here, the communication interface unit 10 also functions as an interface that receives a query for a similar travel route user search and transmits a search result. Hereinafter, the device 1 will be described as a communication equipment device.

[Base station location information management unit 112]
Base station location information management section 112 (FIG. 2) stores a base station identifier and base station location information in association with each other.

  FIG. 3 is a table of base station position information. For each base station identifier, latitude / longitude base station position information is associated. According to FIG. 3, it is understood that the base station 0001 is installed at a position where the latitude is 35.845 degrees and the longitude is 139.502 degrees. Further, it is understood that the base station 0003 is installed at a position of latitude 35.845 and longitude 139.504. Such base station location information may be stored in advance in the base station location information management unit 112, or may be appropriately acquired from each base station 3 via the communication interface unit 10. Also good.

[Communication history collection unit 111]
The communication history collection unit 111 (FIG. 2) collects communication records (communication history) such as date and time and base station identifier from the base station 3 for each mobile terminal 2 subordinate to the base station 3.

FIG. 4 is a communication history table. The communication history is a series of communication records (logs) when the base station 3 accepts communication from the mobile terminal 2. In each communication record, “date and time” and “base station identifier” are associated with each “terminal identifier” (address, telephone number, identification number, etc.) of the mobile terminal 2.
Communication record (terminal identifier, date / time, base station identifier)
According to the communication history of FIG. 4, the mobile terminal 00001 is communicating with the base station 0003 on June 15, 2010, 17:54:50. Further, the portable terminal 00001 is communicating with the base station 0001 on June 15, 2010 at 17:57:00.

[Position information history generation unit 113]
The location information history generation unit 113 (FIG. 2) uses the base station location information management unit 111 to generate a communication history that further associates base station location information corresponding to the base station identifier for each communication record (log). And accumulate.

  FIG. 5 is a communication history table in which base station position information is associated with communication records. The table of FIG. 5 is obtained by replacing the base station identifier portion of the table of FIG. 4 with the latitude / longitude information of the base station of FIG. According to the communication history of FIG. 5, it is understood that the portable terminal 00001 communicated with the base station at latitude 35.845 and longitude 139.504 on June 15, 2010, 17:54:50. Further, it is understood that the portable terminal 00001 communicated with a base station having a latitude of 35.845 and a longitude of 139.502 on June 15, 2010, 17:57:00.

[Time window dividing unit 121]
The time window dividing unit 121 (FIG. 2) divides the communication history output from the position information history generating unit 113 by a predetermined time window (time interval). Specifically, a plurality of communication records constituting the communication history are classified by collecting those belonging to a predetermined time window. The time window is determined by the time width T and the shift width S. The time window of the shift width S is the start time delayed by S. Therefore, when T> S, the time window is overlapped by TS.

  The time width T is a parameter that determines “with which time width to determine stay or stay”. Here, it is difficult to determine stay / movement unless there is a predetermined number or more of position information within the time window. Therefore, it is preferable that the time width T is determined based on the number of communication records (logs) in the communication history so that position information of a predetermined number or more is included in the time window as much as possible.

  Further, when the shift width S is shortened, the time resolution of the stay time section is increased. On the other hand, if the width is increased, the number of time windows increases, and the amount of calculation increases. Therefore, the shift width S is preferably determined according to the time resolution and processing time required by the application.

  The time window dividing unit 121 sets two types of time windows in order to sort the determination of “short stay” and “stay”. Among these, in the determination of “short-term stay”, the actual short-term stay time, for example, the suspension time of the moving means such as a train and the stop time are taken into consideration, and the first time width having a shorter time width (first time width) is taken into consideration. Set 1 time window. Here, the first time width T may be set in consideration of the number of pieces of position information included in the time window. Further, as the first time width T, a distribution relating to the length of time of short stay for each area is acquired, and the average, mode or N quantile (0 ≦ N ≦ 1) of the same time length is used. Also good. In the following example, a first time window with T = 5 minutes and S = 2 minutes 30 seconds is employed.

  On the other hand, in the determination of “stay”, a second time window having a time width (second time width) longer than the first time width is set so that a certain number of position information is included in the time window. . In the following example, a second time window with T = 20 minutes and S = 10 minutes is employed.

  FIG. 6 is a table showing the start time and end time of the first time window (T = 5 minutes and S = 2 minutes 30 seconds). According to FIG. 6, the time window 1 is 17:50:00 to 17:54:59, and T = 00: 05: 00. The time window 2 is 17:52:30 to 17:57:29, and it can be seen that S = 00: 02: 30 with respect to the time window 1.

  FIG. 7 is a table in which each row (communication record) in FIG. 5 is classified for each first time window in FIG. 6 for each portable terminal. In the table, only those having four or more points (position information) in the time window are recorded. According to FIG. 7, for the mobile terminal 00001, five communication records belong to the first time window 12, and four communication records belong to the first time window 13.

  FIG. 8 is a table showing the start time and end time of the second time window (T = 20 minutes and S = 10 minutes). According to FIG. 8, the time window 1 is 17:50:00 to 18:09:59, and T = 00: 20: 00. The time window 2 is 18:00:00 to 18:19:59, and it can be seen that S = 00: 10: 00 with respect to the time window 1.

  FIG. 9 is a table in which each row (communication record) in FIG. 5 is classified for each second time window in FIG. 8 for each portable terminal. According to FIG. 9, for the mobile terminal 00001, five communication records belong to the second time window 1 and six communication records belong to the second time window 2.

[Stay determination unit 122]
The stay determination unit 122 (FIG. 2) determines, for each mobile terminal 2, whether or not it is “stay” based on a probability distribution of positions based on a plurality of base station position information for each second time window. The “stay place” is determined based on the probability distribution of the position in the second time window determined as “stay”. Specifically, for each second time window, if the position probability distribution based on the plurality of base station position information has unimodality, it is determined as “stay”, and if it does not have unimodality, “ "Move".

  Here, unimodality means that the probability distribution of the position has a shape of one mountain (forms one peak). That is, when the probability distribution of the base station position belonging to one second time window (for example, 20 minutes) has a unimodality, it can be determined that this time interval is “stay”. On the other hand, if the probability distribution of the base station position has a plurality of mountain shapes (a plurality of peaks), it is assumed to have multimodality. When there is multimodality for one second time window, it can be determined that this time interval is “moving”.

  The stay determination unit 122 sets a condition regarding the minimum number of points (base station position information number) necessary for the stay determination, and may determine that it is not “stay” in the second time window that is less than the minimum number of points. Good. Hereinafter, the minimum number of points is 4.

  FIG. 10 is a flowchart for determining whether or not there is unimodality.

  According to FIG. 10, the stay determination unit 122 performs representative point calculation processing to determine whether the probability distribution of the location information of the base station has unimodality. As a result, if the number of types of representative points is one, it has unimodality, so it is determined as “stay”, and if there are multiple types of representative points, it has multimodality. It is determined that “move” is in progress.

The representative point calculation process is executed by the following steps S1 to S5 for a plurality of position information included in each second time window.
(S1) An arbitrary point (position information) is set as the first center point.
(S2) The center of gravity is calculated using a point (position information) included in a circle having a radius of a first threshold (for example, 2 km) with the center point as the center.
(S3) Next, it is determined whether or not the difference (change amount) between the calculated center of gravity and the current center point is equal to or less than a second threshold value (for example, 100 m).
(S4) When it is determined to be false in S3, the center of gravity is set as a new center point. Next, the process returns to S2 again, and is repeated until the amount of change falls below the second threshold.
(S5) If it is determined to be true in S3, the center of gravity (the converged point) is set as the representative point.

  Finally, for each second time window, when the number of converged representative points is one, it is determined as “stay”, and when there are a plurality of representative points, it is determined as “move”. Here, with respect to the second time window, when all the representative points fall within the circumferential region having the radius of the first threshold value or the second threshold value, the number of types of converged representative points is one. It is also preferable.

  Hereinafter, for the second time window 1 in FIG. 9, the stay / movement is specifically determined using the representative point extraction process.

(Determination of stay movement in the second time window 1)
(A) The first centroid about the point (35.845, 139.504) of the second time window 1 is calculated. Since the distances between the points of the second time window 1 (35.845, 139.504) and the other points included in the second time window 1 are all within 2 km, the average of all the other points is taken (35.845, 139.503). )
(B) Calculate the second centroid for the second time window 1 point (35.845, 139.504). In the second centroid calculation, the center point is (35.845, 139.503). Since the distance between this center point and other points included in the second time window 1 is all within 2 km, the average of all other points is (35.845, 139.503).
(C) Since the results of the first and second centroid calculations are the same and the variation is 100 m or less, the representative point of the second time window point (35.845, 139.504) is (35.845, 139.503). .
(D) Similarly, when representative points are calculated for each of the other points in the second time window 1, all the representative points are (35.845, 139.503).
(E) As described above, the calculated representative points are (35.845, 139.503) for all points, and the number of types of representative points is 1. Therefore, the second time window 1 is determined to be “stay”. .

  In addition, it is also preferable that the stay determination unit 122 determines whether or not each of the plurality of base station position information has unimodality by using kernel density estimation for each second time window. . The kernel density estimation is a method of extrapolating data of a sample of a certain population in order to estimate a probability density function of a random variable (for example, Tomokazu Nomura, “Kernel density estimation”, [online], [2013] Search August 8, 2012], Internet <URL: http://www.econ.kobe-u.ac.jp/~nomura/lecture/11f/kd.pdf>). According to this method, a hump that becomes a peak can be derived using a kernel function.

  Next, the stay determination unit 122 collects position information of the second time window determined as “stay”. The number of representative points of the second time window determined as “stay” is one. FIG. 11 is a table showing the second time window determined as “stay” and the corresponding representative point position information. According to FIG. 11, the second time windows 1, 2 and 6 are determined to be “stay”.

  The stay determination unit 122 further sets the centroid of the plurality of pieces of position information of each second time window determined as “stay” as “stay place cluster” (“stay place”). Specifically, this center of gravity is used as a representative point, and a set of representative points close to each other is defined as one “stay place cluster”.

  As shown in FIG. 10, the position clustering process for forming this “stay place cluster” also executes the representative point calculation process in the same manner as the process in the stay movement determination described above. Specifically, for the position information of the representative point extracted in each second time window, a centroid is calculated from a given center point within a first threshold range, and the difference (change) between the centroid and the center point is calculated. (Quantity) is repeated until it falls within the second threshold value, and the convergence value (representative point at the representative point in FIG. 11) of each position finally obtained is used as the representative point of the “stay place cluster” for each second time window. And

  FIG. 12 is a table showing the representative point position information of the “stay place cluster” calculated based on the representative point position information for each second time window shown in FIG. 11. In the representative point calculation process in the table, the first threshold value is 2 km and the second threshold value is 100 m. According to FIG. 12, it is understood that one “stay place cluster” is formed for the second time windows 1 and 2, and the representative points of this “stay place cluster” are latitude 35.845 and longitude 139.503. In the second time window 6, it is understood that the positions of latitude 35.862 and longitude 139.515 are representative points of another “stay place cluster”.

  Thus, by using the second time window set appropriately, the “stay” state and “stay place” of the user are estimated using the base station position information without depending on the positioning function of the mobile terminal. It becomes possible to do.

[First short-term residence determination unit 123]
The first short-term residence determination unit 123 (FIG. 2)
(A) For each first time window, it is determined whether or not “staying” based on a probability distribution of positions based on a plurality of base station position information, and the position in the first time window determined as “staying” Based on the probability distribution of
(B) There is a first time window having a different “staying place” before or after a predetermined time or more of a predetermined number of consecutive first time windows determined to be “staying”. In addition, when the “stay place” of the continuous first time window and the different “stay place” are separated by a predetermined distance or more, the continuous first time window is determined as “short stay”. .

  Specifically, the first short-term stay determination unit 123 first determines “stay” if the probability distribution of the positions based on the plurality of base station position information is unimodal for each first time window. If there is no unimodality, it is determined that it is not “staying”. The processing flow for the “stay” determination is the same as the processing flow for the “stay” determination described with reference to FIG. That is, in the “stay” determination processing flow shown in FIG. 10, the first time window is used instead of the second time window, and the “stay” determination is used instead of the “stay” determination. This is the processing flow for determining “stay”.

  FIG. 13 is a table showing the first time window determined as “staying” and the corresponding representative point position information. According to FIG. 13, for the mobile terminal 00001, the first time windows 13, 17 and 18 are determined to be “staying”. For the portable terminal 00002, the first time windows 13, 15, 16, and 17 are determined to be “staying”.

  The first short-term stay determination unit 123 further sets the center of gravity of the plurality of pieces of position information of each first time window determined as “stay” as “stay place”. Specifically, this center of gravity is used as a representative point, and a set of representative points that are close to each other is used as one “staying place”. Also in the position clustering process for forming the “stay place”, the representative point calculation process is executed in the same manner as the above-described stay movement determination process (FIG. 10).

  Next, the first short-term stay determination unit 123 has different “stay places” before or after a predetermined time or more of the predetermined time period with respect to a predetermined number of consecutive first time windows determined as “stay”. When the first time window exists and the “residence place” of the continuous first time window is separated from the different “stay place” by a predetermined distance or more, the continuous first window The time window is determined as “short dwell”.

  In the following examples, “determined” as “determined” in two or less (ie, two) consecutive first time windows and different “residence” before 2 minutes 30 seconds or more after retention or 2 minutes 30 seconds or more. It is determined as “short-time stay” when it stays at “ground” and the distance between “stay ground” is 2 km or more.

  FIG. 14 is a table showing the first time window determined to be “short stay” and the corresponding “stay place” representative point. According to FIG. 14, with respect to the portable terminal 00001, the first time window 13 and the first time windows 17 and 18 are determined to be “short stay”. Further, for the mobile terminal 00002, the first time window 13 and the first time windows 15 and 16 are determined to be “short-term stay”.

  In this way, by using the appropriately set first time window, it is possible to estimate the user's “short-term residence” using the base station position information without depending on the positioning function of the mobile terminal. It becomes.

[Base station information collection unit 124]
The base station information collection unit 124 (FIG. 2), based on the probability distribution of the “staying place” of the first time window determined as “short staying” for a plurality of mobile terminals 2 (multiple users), The center of gravity is calculated as a “short-time staying place” (retaining place representative point).
A base station list in which (a) the base station position information is associated with (b) the connection probability for the base station is generated. Here, when the collection destination of the information related to “stay” is limited to one portable terminal 2 (one user), there are few cases having a certain number of connection times or more in “short stay place” and there is a limit to data collection. Although it arises, if it collects from the several portable terminal 2 (plural users), it will become possible to collect a fixed number of cases substantially reliably.

  FIG. 15 is a schematic diagram showing how information related to “stay” is collected from a plurality of mobile terminals 2, and a table showing calculation results of staying place representative points in “short staying place”. FIG. 16 is a table showing a base station list of short staying places.

  According to FIG. 15, short-time staying places 1 to 3 are listed as short-time staying places. In FIG. 16, for example, for the short-time staying place 1, three corresponding base station position information is listed, and a connection probability is associated with each position information (base station). For example, in the base station of the location information (35.848, 139.506) of the short staying place 1, the number of connections is 2, but since the total number of connections in the short staying place 1 is 2 + 2 + 4 = 8, the connection probability is 2 / 8 = 0.250.

[Second short time residence determination unit 125]
When there is a base station included in the base station list (FIG. 16) of the short-time stay location for the acquired communication history, the second short-term stay determination unit 125 (FIG. 2) A short-time staying place history in which “short-time staying place” corresponding to the base station list is associated with the position information is generated. That is, this base station identifier is replaced with a short staying place number. Note that the base station list of short-time staying locations used for the determination may be limited to base stations having a connection probability of “short-time staying place” or higher. This embodiment is not limited.

  FIG. 17 is a table showing the results of the second short-term stay determination.

  According to FIG. 17, the mobile terminal 00001 “short-term stay” from the disclosure time 18:21:10 to the end time 18:21:40, and then discloses the short-time residence place 2 “Short Stay” from 18:28:20 to End Time 18:30:00, then “Short Stay” for Short Stay 3 from Disclosure Time 18:33:10 to End Time 18:35:20 It ’s understood that As described above, according to the second short-term stay determination result, it is possible to acquire a short-term stay history in which the mobile terminal 2 has “short-term stay”.

[Stay staying time series data generation unit 126]
The stay residence time series data generation unit 126 (FIG. 2) sets the “stay place” and / or “short stay place” determined for each portable terminal 2 based on the generated short stay place history. Generate stay residence time series data arranged in series. The “stay place” data in which the start time and the end time are associated with each other is stored in the base station information collection unit 124 that has input the “stay place” determined by the stay determination unit 122 for each “stay place”. It is acquired by generating a base station list in which information is associated.

  FIG. 18 is a schematic diagram showing how the mobile terminals 00001 and 00002 progress while performing “stay” and “short stay”, and a table showing the generated stay stay time series data.

  According to FIG. 18, it is understood how the portable terminal 00001 reaches from the stay place 1 through the short stay place 1 to the short stay place 3. On the other hand, as for the portable terminal 00002, the background from the short staying place 1 to the staying place 2 through the short staying place 2 is understood. Both mobile terminals stayed in the short staying place 1 from the start time 18:20:00 to the end time 18:24:59, for example, staying at the same station in this time range. I understand that.

[Short Staying Location Transition Model Building Unit 127]
The transition model construction unit 127 (FIG. 2) for short staying places extracts and extracts transition patterns between “short staying places” that occur at a predetermined frequency or more from the staying time series data as shown in FIG. For each transition pattern between the “short-time staying places”, a transition model between “short-time staying places” in which the appearance probability of the transition pattern is associated is constructed. Specifically, “short-time staying places” that appear with a transition probability equal to or higher than a predetermined threshold before and after “short-time staying place” are detected.

  FIG. 19 is a table showing a transition model between “short-time staying places”.

  In FIG. 19, transition patterns that appear with a probability of 0.5 or more are extracted. Here, as shown in FIG. 18, since there are two types of staying residence time series data, an appearance probability of 0.5 or more corresponds to the appearance of transition patterns in one or more types of time series data. Further, in FIG. 19, the average required time is recorded in association with each transition pattern. For each transition pattern, this average required time is the difference between the intermediate time between the start time and end time at the transition source “Short Stay Location” and the intermediate time between the start time and end time at the transition destination “Short Stay Location”. It is calculated by calculating and taking the average.

[Transition model construction unit 128 between staying place and short staying place]
The transition between the staying place and the short staying place model construction unit 128 (FIG. 2) generates the transition between the staying place and the short staying place generated from the stay staying time series data as shown in FIG. A pattern is extracted, and for each extracted transition pattern between “stay place / short-time staying place”, a transition model between “stay place / short-time staying place” is constructed in association with the appearance probability of the transition pattern. Specifically, “short-time staying places” that appear with a transition probability equal to or higher than a predetermined threshold before and after “staying place” are detected.

  FIG. 20 is a table showing a transition model between “stay place and short stay place”.

In FIG. 20, transition patterns that appear with a probability of 0.5 or more are extracted. Here, as shown in FIG. 18, there are two types of staying residence time-series data, and therefore, an appearance probability of 0.5 or more corresponds to the appearance of transition patterns in one or more types of time-series data. Further, in FIG. 19, the average required time is recorded in association with each transition pattern. For each transition pattern, this average required time is the intermediate time between the start time and end time at the transition destination / original “short staying place”, and the start time and end time at the transition source / destination “stay place”. The difference is calculated by calculating the difference from the one closer to the intermediate time and taking the average.
[Example of building a transition model between short-term staying places]

  Here, a simple example of construction of a transition model between short staying places will be described. For the staying and staying time series data of a plurality of mobile terminals, when another short staying place 2 is detected after a short staying place 1 within a predetermined time (for example, 5 minutes), the transition is recorded, Based on the number of transitions in all transitions recorded for the time staying place 1, the transition probability from the short staying place 1 to the short staying place 2 is calculated. The total transition probabilities associated with the short staying place 1 are 1. In the following, a transition model between short staying places is constructed for cases where A to G exist as short staying places.

  FIG. 21 is a table showing a transition model between short-time staying places A to G.

According to FIG. 21, when the short staying place 1 is A for a plurality of portable terminals, the probability that the short staying place 2 detected after the short staying place 1 is B is 0.9. Yes, the probability of being E is 0.1. Therefore, it is understood that the mobile terminal transitions to the short staying place B with a probability of 90% following the short staying place A, and transitions to the short staying place E when the remaining 10%.
[Example of construction of transition model between staying place and short staying place]

  Next, a simple example of construction of a transition model between a stay place and a short stay place will be described. About the stay residence time series data for each portable terminal, when a short stay location 2 is detected following a stay location 1 within a predetermined time (for example, 5 minutes), the transition is recorded, and the stay location 1 is recorded. Based on the number of transitions in all transitions, the transition probability from the staying place 1 to the short staying place 2 is calculated. The total transition probability from stay 1 is 1. On the other hand, a transition model from the short staying place 1 to the staying place 2 is similarly constructed. In this case, the total probability of transition to the place of stay 2 is 1. In the following, a transition model between short staying places is constructed for cases where α and β exist as staying places and A to G exist as short staying places.

  FIG. 22 is a table showing a stay model / short-time stay place transition model for stay places α and β and short stay places A to G.

  According to FIG. 22, when the stay place 1 is α for a certain mobile terminal, the probability that the short stay place 2 detected after the stay place 1 is A is 0.9 and is E. The probability is 0.1. Therefore, it is understood that the mobile terminal transits to the short staying place A with a probability of 90% following the staying place α, and transitions to the short staying place E when the remaining 10%.

  On the other hand, for a certain mobile terminal, if the stay destination 2 of the transition destination is β, the probability that the short-time staying place 1 detected before the stay place 2 is D is 0.9 and the probability that it is G Is 0.1. Therefore, it is understood that 90% of the transition source to the stay place β is the short staying place D and the remaining 10% is the short staying place G.

[Short Stay Location Interpolation Unit 131]
The short-time staying place interpolation unit 131 (FIG. 2) creates a “short-time staying place” transition model (FIG. 21) and “ Based on the transition model between the “stay place / short stay place” (FIG. 22), the “short stay place” missing in the time series data is interpolated. Specifically, there is a high probability of appearance between “stay place” and “short stay place” or “short stay place” that do not exist in the stay stay time series data but appear in time series If there is a “short-time staying place” that can appear in the above, the “short-time staying place” is interpolated.

  Incidentally, even if the mobile terminal 2 that is moving is located under any of the base stations 3, the necessity for such interpolation is not necessarily due to the presence of radio wave obstacles, communication congestion, congestion, etc. This is because a connection with the base station 3 cannot always be established. If communication between the portable terminal 2 and the base station 3 is not possible, the communication history (FIG. 5) in which the base station position information is associated with the communication record cannot be acquired, and the “stay place” and “short time” It is also impossible to estimate “residence”.

  Here, the short-time staying place interpolation unit 131 has a case where adjacent “short-time staying places” in the stay staying time series data or adjacent “staying place” and “short-time staying place” are separated by a predetermined time or more. It is also preferable to estimate that there is a missing “short-time staying place” between these “short-time staying places” or between these “staying place” and “short-time staying place”.

  In addition, when there are a plurality of candidates for “short-time staying place” that are missing in the time series, the short-time staying place interpolation unit 131 has the highest likelihood based on the appearance probability of the transition pattern including the candidate in between. It is also preferable to interpolate with the candidate that becomes higher.

  FIG. 23 is a table for explaining an example of “short-time staying place” interpolation.

  In FIG. 23, transition patterns 1 to 5 of “stay place” and “short stay place” are listed. In any transition pattern, the movement starts from the staying place α, passes through a plurality of short-time staying places, and ends moving at the staying place β. For each pattern, when adjacent “stay place” and “short-time stay place” and adjacent “short-time stay place” are separated by 5 minutes or more (as a predetermined time), Complements the “short-time staying place”.

As shown in FIG. 23, in pattern 1, the short-time staying place A and the short-time staying place C are separated by 5 minutes or more, so one short-time staying place <1> is interpolated between them. Here, according to the transition model between “short-time staying places” in FIG.
A → B → C
Only. That is, the transition destination from A is B and E, but the transition source to C is only B. Therefore, only B can be interpolated between A and C. Accordingly, the short staying place B is determined as the short staying place <1>.

Next, in pattern 2, since the short staying place B and the short staying place D are separated by 5 minutes or more, one short staying place <2> is interpolated between them. According to the transition model between “short-term staying places” in FIG. 21, the transition from B to D is
B → C → D and B → F → D
These are two. Here, the likelihood of the transition pattern which employ | adopted said each transition is calculated and compared, and a transition with higher likelihood is employ | adopted. In the present embodiment, the likelihood is calculated as a product of transition probabilities of transition elements (for one arrow) constituting the transition pattern. The transition probability of each transition element is a value determined by the transition model between the “short-time staying place” (FIG. 21) and the transition model between the “staying place / short-time staying place” (FIG. 22).

The likelihood P (B, C, D) when B → C → D is adopted is
(1) P (B, C, D) = P (α, A) ・ P (A → B) ・ P (B → C) ・ P (C → D) ・ P (D → β)
The likelihood P (B, F, D) when B → F → D is adopted is
(2) P (B, F, D) = P (α, A) ・ P (A → B) ・ P (B → F) ・ P (F → D) ・ P (D → β)
It becomes. Here, P (X, Y) is a transition probability from the stay place (short stay place) X to the short stay place (stay place) Y.

The difference between the formulas (1) and (2) is the part of P (B → C) · P (C → D) and the part of P (B → F) · P (F → D). Compare the values of these parts.
(1a) P (B → C) · P (C → D) = 0.5 × 0.9 = 0.45
(2a) P (B → F) ・ P (F → D) = 0.5 × 0.2 = 0.10
Therefore, the transition pattern adopting B → C → D has a higher likelihood. Accordingly, the short staying place C is determined as the short staying place <2>.

Next, in pattern 3, a short-time staying place is complemented between A and D. According to the transition model between “short-time staying places” in FIG. 21, the transition from A to D through a short staying place is
A → B → C → D, A → B → F → D, and A → E → F → D
It is three. Again, the likelihood of the transition pattern employing each of the above transitions is calculated and compared, and a transition with a higher likelihood is employed.

Likelihood P (A, B, C, D), P (A, B, F, D) when A → B → C → D, A → B → F → D and A → E → F → D ) And P (A, E, F, D) are
(3) P (A, B, C, D) = P (α, A) ・ P (A → B) ・ P (B → C) ・ P (C → D) ・ P (D → β)
(4) P (A, B, F, D) = P (α, A) ・ P (A → B) ・ P (B → F) ・ P (F → D) ・ P (D → β)
(5) P (A, E, F, D) = P (α, A) ・ P (A → E) ・ P (E → F) ・ P (F → D) ・ P (D → β)
It becomes.

When calculating and comparing values of different parts between the formulas (3) to (5),
(3a) P (A → B) ・ P (B → C) ・ P (C → D) = 0.405
(4a) P (A → B) ・ P (B → F) ・ P (F → D) = 0.090
(5a) P (A → E), P (E → F), P (F → D) = 0.020
Therefore, it can be seen that the transition pattern adopting A → B → C → D has the highest likelihood. Therefore, the short-time residences B and C are determined as the short-time residences <3> and <4>, respectively.

Next, in the pattern 4, since the staying place α and the short staying place B are separated by 5 minutes or more, one short staying place <5> is interpolated between them. According to the transition model between “stay place and short stay place” in FIG. 22 and the transition model between “short stay place” in FIG. 21, the transition from α to B is
α → A → B
There is only one. Accordingly, the short staying place A is determined as the short staying place <5>.

Finally, in Pattern 5, since the short staying place C and the staying place β are separated by 5 minutes or more, the short staying place <6> is supplemented between C and β. According to the transition model between “stay place and short stay place” in FIG. 22, the transition from C to β is
C → D → β and C → G → β
These are two. Again, the likelihood of the transition pattern employing each of the above transitions is calculated and compared, and a transition with a higher likelihood is employed.

Likelihoods P (C, D, β) and P (C, G, β) when C → D → β and C → G → β are adopted respectively.
(6) P (C, D, β) = P (α, A) ・ P (A → B) ・ P (B → C) ・ P (C → D) ・ P (D → β)
(7) P (C, G, β) = P (α, A) ・ P (A → B) ・ P (B → C) ・ P (C → G) ・ P (G → β)
It becomes.

When calculating and comparing values of different parts between the equations (6) and (7),
(6a) P (C → D) · P (D → β) = 0.81
(7a) P (C → G) · P (G → β) = 0.01
Therefore, it can be seen that the transition pattern adopting C → D → β has a higher likelihood. Therefore, the short staying place D is determined as the short staying place <6>.

  If the short staying place interpolation unit 131 determines that it is not necessary to interpolate the “short staying place” in the stay staying time series data, the stay staying time series data, and therefore the result of the second short staying judgment. (FIG. 17) is output as it is.

[User search section]
The user search unit 132 (FIG. 2) uses the stay stay time series data after the above short-term stay location interpolation processing related to a predetermined mobile terminal and the stay stay time after the short stay stay location interpolation processing related to another mobile terminal. Based on the longest matching partial sequence length with the sequence data, a mobile terminal that takes a movement route similar to the movement route of the predetermined mobile terminal is searched. Here, the longest matching partial sequence length is, for example, a common portion C → D → E in A → B → C → D → E and A → C → D → E → F in two time series data to be compared. Is defined as the length of the portion where the time series of the “short-time staying place” coincide.

Specifically, the user search unit 132 calculates the similarity between the two portable terminals 2.
(8) (Similarity) =
(Longest matching partial sequence length) / (longer sequence length of time series data)
As a result, it is possible to search for portable terminals having this similarity equal to or greater than a predetermined threshold. Here, the denominator (the longer sequence length of the time series data) is the longer sequence length of the staying residence time series data after the interpolation processing related to the two mobile terminals 2. The similarity is normalized by dividing by this (the longer sequence length of the time series data). In addition, the user search unit 132 sets the longest matching subsequence for a time interval between time intervals corresponding to the “stay places” separated from each other by a predetermined time or more among the search target time intervals designated by the user. It is also preferable to investigate the length.

  As an example, the designated time interval is 18:20 to 18:35, the designated mobile terminal is the mobile terminal 00001, which is similar to the movement route of the mobile terminal 00001 in this designated time interval Search for one or more users on the travel route. In this search, the similarity between the portable terminal 00001 and another user is calculated, and a user whose similarity is a predetermined threshold = 0.8 or more is searched.

Hereinafter, it is assumed that the result of the second short-term stay determination in FIG. 17 is the result after the correction processing described above. According to this short-term residence determination result, the portable terminal 00001 and the portable terminal 00002 are as a common time series of “short-term residence place” in the specified time section 18:20 to 18:35,
Short staying place 1, Short staying place 2, Short staying place 3
have. Therefore, (longest matching partial sequence length) = 3. Further, since the time series data lengths of the portable terminal 00001 and the portable terminal 00002 are both 3, (the longer series length of the time series data) = 3. Therefore, (similarity) = 3/3 = 1 is obtained from equation (8). Since the predetermined threshold value of similarity counterintelligence is 0.8, the similarity between the mobile terminal 00001 and the mobile terminal 00002 is equal to or greater than the predetermined threshold value. Therefore, the portable terminal 00002 is searched for as a portable terminal of the user who takes a movement route similar to the movement route of 18:20 to 18:35 for the user of the portable terminal 00001.

  As described above in detail, according to the apparatus, program, and method of the present invention, the mobile terminal can be obtained using the base station position information that can be acquired by the communication carrier equipment without relying on the positioning function of the mobile terminal. It is possible to search for a user who takes a movement route similar to the movement route of the possessed user. Here, the base station location information generally has a coarse spatial granularity and the time interval is not constant. However, even if such base station location information is used, the user's “stay place” and / or “short stay” It is possible to calculate stay residence time series data in which “land” is arranged in time series. In particular, according to the present invention, while avoiding a processing load for positioning on a mobile terminal, information on the staying time series data of a user, and further information on estimating a user who takes a similar movement route is transmitted to a communication business. Can be obtained only by the person in charge.

  Furthermore, according to the present invention, the user's travel route is estimated with higher accuracy despite the presence of a deficiency in the stay residence time series data that can normally occur. As a result, it is possible to more reliably search for a user who takes a similar movement route.

  Here, as a service provided by searching for a user who takes a movement route similar to the user's movement route, for example, for a user group possessing a mobile terminal and having a movement route similar to each other, Online information such as advertisements and coupons according to surrounding information can be distributed. Thereby, the improvement of the advertising effect of the distribution information is expected. Further, it is possible to provide judgment data for designing and constructing a suitable wireless communication network, such as adding a base station in an area where movement routes are concentrated.

  Various changes, modifications, and omissions of the various embodiments of the present invention described above within the scope of the technical idea and the viewpoint of the present invention can be easily made by those skilled in the art. The above description is merely an example and is not intended to be a limitation. The present invention is limited only by the claims and the equivalents thereof.

1 Communication equipment (similar travel route user search device)
DESCRIPTION OF SYMBOLS 10 Communication interface part 111 Communication history collection part 112 Base station position information management part 113 Position information history storage part 12 Time series data generation part 12
121 time window division unit 122 stay determination unit 123 first short stay determination unit 124 base station information collection unit 125 second short stay determination unit 126 stay stay time series data generation unit 127 short time stay transition model construction Unit 128 Location / Short Stay Location Transition Model Construction Unit 131 Short Stay Location Interpolation Unit 132 User Search Unit 2 Mobile Terminal 3 Base Station

Claims (11)

An apparatus for searching for a user who takes a movement route similar to the movement route of a user possessing a mobile terminal,
Location information history storage means for storing a communication history including a plurality of communication records in which date and time related to communication and base station location information are associated with each portable terminal;
For each portable terminal, a “short-time staying place” is determined based on a probability distribution of positions based on position information of a plurality of base stations in the first time window in the communication history, and is longer than the first time window. A “stay place” is determined based on a probability distribution of positions based on position information of a plurality of base stations in a second time window having a time width, and the determined “stay place” and / or “short stay place” is determined. Time series data generation means for generating stay residence time series data arranged in time series,
A transition pattern between “short-time staying places” that occurs at a predetermined frequency or more is extracted from the stay residence time series data, and the transition pattern appearance probability is associated with each transition pattern between the “short-time staying places”. A transition model construction means for constructing a transition model between short-term staying places,
The transition pattern between “stay place / short-time stay place” that occurs at a predetermined frequency or more is extracted from the stay stay time series data, and the transition pattern appears for each transition pattern between the “stay place / short stay place” A transition model construction means for building a transition model between a staying place and a short staying place that associates a probability,
Based on the transition model between the “short-term staying place” and the transition model between the “staying place / short-time staying place”, the “short-term staying” missing in the time series Short-term staying ground interpolation means for interpolating the ground,
Based on the longest matching partial sequence length between the stay and stay time series data after the interpolation process related to a predetermined mobile terminal and the stay and stay time series data after the interpolation process related to another mobile terminal, An apparatus comprising user search means for searching for a portable terminal that takes a movement route similar to the movement route of the portable terminal. The time series data generating means includes
A time window dividing means for dividing the communication history by the first time window and the second time window;
For each of the first time windows, it is determined whether or not “staying” based on a probability distribution of positions based on a plurality of base station position information, and the position of the first time window determined as “staying” is determined. Based on the probability distribution, the “staying place” is determined, and the “first staying place” that is different from before or after the predetermined time of the staying with respect to a predetermined number of first time windows determined to be “staying” is different. When there is a first time window and the “stagnation place” of the continuous first time window and the different “stagnation place” are separated by a predetermined distance or more, the continuous first time period First short-term dwell determination means for determining the window as “short dwell”;
Based on the probability distribution of “stay place” in the first time window determined to be “short stay” related to a plurality of mobile terminals, “short stay place” is determined, and for each “short stay place” Base station information collection means for generating a base station list in which base station position information is associated;
For the communication history, when there is a base station included in the base station list, a short time in which the base station position information of the base station is associated with the “short-time staying place” corresponding to the base station list. A second short-term residence determining means for generating a residence location history;
For each second time window, it is determined whether or not it is “stay” based on a probability distribution of positions based on a plurality of base station position information, and the probability of the position in the second time window determined as “stay” A stay determination means for determining a “stay place” based on the distribution;
Based on the generated short staying place history for each mobile terminal, stay staying that generates stay staying time-series data in which the determined “staying place” and / or “short staying place” is arranged in time series The apparatus according to claim 1, further comprising time series data generation means.   When there are a plurality of candidates for “short-time staying place” that are missing in time series, the short-time staying place interpolation means has the highest likelihood based on the appearance probability of the transition pattern that includes the candidate. However, the apparatus according to claim 1, wherein interpolation is performed using the candidate.   The short staying place interpolation means, when the adjacent "short staying places" in the stay staying time series data or adjacent "stay place" and "short staying place" are separated by a predetermined time or more, 2. It is estimated that there is a missing “short-time staying place” between “short-time staying places” or between the “staying place” and the “short-time staying place”. 4. The apparatus according to any one of items 3.   The user search means sets the similarity between two portable terminals as (longest matching partial sequence length) / (longer sequence length of staying residence time series data after interpolation processing related to the two portable terminals) 5. The apparatus according to claim 1, wherein a mobile terminal having a similarity equal to or higher than a predetermined threshold is searched.   The user search means investigates the longest matching partial sequence length for a time section that is between time sections corresponding to “stay places” that are separated from each other by a predetermined time or more among the specified search target time sections. The apparatus according to any one of claims 1 to 5. The base station information collecting means generates a base station list in which the base station position information of the base station and the connection probability with respect to the base station are associated with each other for each "short staying place",
3. The apparatus according to claim 2, wherein the second short-term residence determination unit selects only a base station having the connection probability equal to or higher than a predetermined value and generates the short-term residence location history.   The time window dividing means sets an average value, a mode value, or an N quantile (0 ≦ N ≦ 1) value of the time length of short-term stay in a predetermined area as a time width of the first time window. An apparatus according to claim 2 or 7, characterized in that A communication equipment device according to any one of claims 1 to 8, connected to a wide area wireless communication network,
Base station location information management means for storing a base station identifier and base station location information in association with each other;
From a base station to which the mobile terminal is connected under control, the mobile station further includes a communication history collecting means for collecting a communication record that associates the date and time of communication for each mobile terminal with the base station identifier of the base station,
The location information history accumulating unit generates the communication history by further associating base station location information corresponding to a base station identifier for each mobile terminal with respect to the communication record. A program for causing a computer mounted on a device to function to search for a user who takes a movement route similar to the movement route of a user possessing a mobile terminal,
Location information history storage means for storing a communication history including a plurality of communication records in which date and time related to communication and base station location information are associated with each portable terminal;
For each portable terminal, a “short-time staying place” is determined based on a probability distribution of positions based on position information of a plurality of base stations in the first time window in the communication history, and is longer than the first time window. A “stay place” is determined based on a probability distribution of positions based on position information of a plurality of base stations in a second time window having a time width, and the determined “stay place” and / or “short stay place” is determined. Time series data generation means for generating stay residence time series data arranged in time series,
A transition pattern between “short-time staying places” that occurs at a predetermined frequency or more is extracted from the stay residence time series data, and the transition pattern appearance probability is associated with each transition pattern between the “short-time staying places”. A transition model construction means for constructing a transition model between short-term staying places,
The transition pattern between “stay place / short-time stay place” that occurs at a predetermined frequency or more is extracted from the stay stay time series data, and the transition pattern appears for each transition pattern between the “stay place / short stay place” A transition model construction means for building a transition model between a staying place and a short staying place that associates a probability,
Based on the transition model between the “short-term staying place” and the transition model between the “staying place / short-time staying place”, the “short-term staying” missing in the time series Short-term staying ground interpolation means for interpolating the ground,
Based on the longest matching partial sequence length between the stay and stay time series data after the interpolation process related to a predetermined mobile terminal and the stay and stay time series data after the interpolation process related to another mobile terminal, A program that causes a computer to function as user search means for searching for a portable terminal that takes a movement route similar to the movement route of the portable terminal. A method of searching for a user who takes a movement route similar to the movement route of a user possessing a mobile terminal using an apparatus,
The apparatus has location information history storage means for storing a communication history including a plurality of communication records in which date and time related to communication and base station location information are associated with each mobile terminal, and the method includes:
For each portable terminal, a “short-time staying place” is determined based on a probability distribution of positions based on position information of a plurality of base stations in the first time window in the communication history, and is longer than the first time window. A “stay place” is determined based on a probability distribution of positions based on position information of a plurality of base stations in a second time window having a time width, and the determined “stay place” and / or “short stay place” is determined. A first step of generating stay residence time series data arranged in time series,
A transition pattern between “short-time staying places” that occurs at a predetermined frequency or more is extracted from the stay residence time series data, and the transition pattern appearance probability is associated with each transition pattern between the “short-time staying places”. A second step of building a transition model between "short-term staying places";
The transition pattern between “stay place / short-time stay place” that occurs at a predetermined frequency or more is extracted from the stay stay time series data, and the transition pattern appears for each transition pattern between the “stay place / short stay place” A third step of constructing a transition model between “stay place / short-time stay place” in which probabilities are associated;
Based on the transition model between the “short-term staying place” and the transition model between the “staying place / short-time staying place”, the “short-term staying” missing in the time series A fourth step of interpolating the “ground”;
Based on the longest matching partial sequence length between the stay and stay time series data after the interpolation process related to a predetermined mobile terminal and the stay and stay time series data after the interpolation process related to another mobile terminal, And a fifth step of searching for a portable terminal that takes a movement route similar to the movement route of the portable terminal.

Images