JP2014182609A - Classifier construction device, classification apparatus, classifier construction method, classification method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique used for classifying a place where a user frequently stays with higher precision.SOLUTION: A classifier construction device is configured to: generate a cluster of a staying place group of a user using a portable terminal according to positioning information of the portable terminal; calculate a predetermined character value with regard to the generated cluster; refer to positional information a prescribed action base of the user of the portable terminal; and determine whether or not the generated staying place cluster is classified to the prescribed action base. The classifier construction device performs supervised learning according to the calculated character value and a determination result to generate a classifier.

Description

  The present invention relates to a technique for processing positioning information of a mobile terminal.

  A technique for classifying places where a user often stays using positioning information of a portable terminal is known. For example, in Cited Document 1, there is a place where a user frequently stays using information on the total stay time and stay time of a user at a certain place calculated from positioning information acquired by GPS (Global Positioning System). A technique for classifying home or work location is disclosed.

Special table 2012-507760 gazette

  However, simply using information on the total stay time and stay time zone as a condition as in the above-described prior art, for example, what value is the total stay time and stay time zone? It may be difficult to determine appropriate classification criteria for whether to classify properly. Therefore, a method for classifying the place where the user stayed with higher accuracy is desired.

  This invention is made | formed in view of the above, The objective is to provide the technique used in order to classify | categorize the place where a user often stays with higher precision.

  The classifier construction apparatus according to the present invention detects a stay location of a user of the mobile terminal based on positioning information of the mobile terminal, and generates a cluster of the detected stay locations (hereinafter referred to as stay location cluster). Referring to the generated stay place cluster, the calculation means for calculating the value of the feature defined for the stay place cluster, and the position information of the predetermined action base associated with the user of the mobile terminal, and the generation Determination means for determining whether or not the stay destination cluster is classified into the predetermined action base, supervised learning based on the calculated feature value and the determination result, and the feature Classifier generating means for generating a classifier having an explanatory variable and whether or not the stay destination cluster is classified as a predetermined action base is used.

  The classification device according to the present invention detects a place of stay of a user of the mobile terminal based on storage means for storing a classifier and positioning information of the mobile terminal, and the cluster of the detected stay place (hereinafter referred to as a stay place cluster). ) For generating the stay destination cluster, using the calculation means for calculating the value of the feature defined for the stay destination cluster, and using the calculated feature value as an explanatory variable, Classifying means for classifying the stay location cluster generated by the generating means by a stored classifier, and the classifier detects the stay location of the user of the mobile terminal based on the positioning information of the mobile terminal; Generating the detected classifier stay location cluster, calculating the classifier stay location cluster for the generated classifier stay location cluster, Referring to position information of a predetermined action base associated with the user of the mobile terminal, it is determined whether the generated classifier staying place cluster is classified into the predetermined action base, and the calculation is performed. A classifier generated by using supervised learning based on the value of the feature and the determination result, using the feature as an explanatory variable, and whether the destination cluster is classified as a predetermined action base It is characterized by being.

  A classifier construction method according to the present invention is a classifier construction method implemented in an information processing apparatus including a control unit, wherein the control unit is based on positioning information of the mobile terminal, and a place where the user of the mobile terminal stays And generating a cluster of the detected stay place (hereinafter referred to as stay place cluster), and the control unit defines a value of the feature defined for the stay place cluster for the generated stay place cluster A calculating step of calculating the location, and the control unit refers to position information of a predetermined action base associated with the user of the mobile terminal, and the generated stay place cluster is classified as the predetermined action base. A determination step for determining whether or not the control unit performs supervised learning based on the calculated feature value and the determination result, sets the feature as an explanatory variable, Cluster; and a classifier generation step of generating a classifier for the purpose variable whether is classified into a predetermined basing.

  A classification method according to the present invention is a classification method implemented in an information processing apparatus including a control unit and a storage unit that stores a classifier, wherein the control unit is configured to perform the portable operation based on positioning information of a portable terminal. A generation step of detecting a stay location of the user of the terminal and generating a cluster of the detected stay location (hereinafter referred to as stay location cluster), and the control unit defining the stay location cluster with respect to the generated stay location cluster A calculation step of calculating a value of the feature being performed, and a stay location cluster generated by the generation step by the stored classifier using the calculated feature value as an explanatory variable A classifying step, wherein the classifier detects a place of stay of the user of the mobile terminal based on the positioning information of the mobile terminal, and the classifier stay detected Generating a cluster, calculating a value of a feature defined for the stay location cluster for the generated classifier stay location cluster, and referring to position information of a predetermined action base associated with the user of the mobile terminal Determining whether or not the generated classifier stay location cluster is classified into the predetermined action base, and performing supervised learning based on the calculated feature value and the determination result, The classifier is generated by using the feature as an explanatory variable and whether the stay destination cluster is classified as a predetermined action base or not as an objective variable.

  The program according to the present invention is a generating means for detecting a place of stay of a user of the mobile terminal based on positioning information of the mobile terminal and generating a cluster of the detected stay place (hereinafter referred to as a stay place cluster). With respect to the generated stay place cluster, a calculation means for calculating a value of a feature defined for the stay place cluster, referring to position information of a predetermined action base associated with a user of the mobile terminal, the generated Determining means for determining whether or not a stay location cluster is classified as the predetermined action base, performing supervised learning based on the calculated feature value and the determination result, and using the feature as an explanatory variable And a classifier generation unit that generates a classifier having a target variable that determines whether or not the stay cluster is classified as a predetermined action base. And wherein the door.

  A program according to the present invention detects a place of stay of a user of a mobile terminal based on positioning information of a storage means for storing a classifier and a mobile terminal, and a cluster of the detected stay places (hereinafter referred to as a stay place). Generating means for generating a cluster), calculating means for calculating a value of a feature defined for the stay destination cluster for the generated stay place cluster, and using the calculated feature value as an explanatory variable, the storage Classifying means for classifying the stay location cluster generated by the generating means by the generated classifier, wherein the classifier detects the stay location of the user of the mobile terminal based on the positioning information of the mobile terminal, and Generate the detected classifier stay cluster and, for the generated classifier stay cluster, use the feature value defined for the stay cluster. Determining whether or not the generated classifier stay location cluster is classified into the predetermined action base with reference to position information of the predetermined action base associated with the user of the mobile terminal; Supervised learning is performed based on the calculated feature value and the determination result, the feature is used as an explanatory variable, and whether or not the stay location cluster is classified as a predetermined action base is generated as an objective variable. It is made to function as a classification device which is a classifier.

  The program of the present invention can be installed or loaded on a computer through various recording media such as an optical disk such as a CD-ROM, a magnetic disk, and a semiconductor memory, or via a communication network. .

  Further, in this specification and the like, the “unit” does not simply mean a physical configuration, but also includes a case where the functions of the configuration are realized by software. In addition, functions of one configuration may be realized by two or more physical configurations, or functions of two or more configurations may be realized by one physical configuration.

  ADVANTAGE OF THE INVENTION According to this invention, the technique used in order to classify | categorize the place where a user stays well with higher precision can be provided.

It is a block diagram which shows the outline of the hardware constitutions of the information processing apparatus in one Embodiment. It is a block diagram which shows the outline of a function structure of the classifier construction apparatus in one Embodiment. It is a block diagram which shows the outline of a function structure of the classification device in one Embodiment. In one Embodiment, it is a figure which shows the structure of the information memorize | stored in the memory | storage part. It is a flowchart which shows the process of classifier construction in one Embodiment. It is a flowchart which shows the process of the classification | category in one Embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. However, the scope of the invention is not limited to these.

  With reference to FIG. 1, an outline of a hardware configuration of an information processing apparatus used to implement a classifier construction apparatus or a classification apparatus according to an embodiment will be described. The information processing apparatus 10 mainly includes a control unit 11, a communication unit 14, a storage unit 15, a display 16, and an operation unit 17 as a hardware configuration. The control unit 11 mainly includes a CPU (Central Processing Unit) 12 and a memory 13. The information processing apparatus 10 can be configured using a computer such as a server apparatus or a terminal apparatus. For example, the information processing apparatus 10 functions as various function realization means by the CPU 12 executing a predetermined program stored in the memory 13 or the like. To do. The information processing apparatus 10 may be configured by a single information processing apparatus or may be configured by a plurality of information processing apparatuses distributed on a network.

  In the control unit 11, the CPU 12 develops and executes the program stored in the storage unit 15 and the like in the memory 13, thereby controlling operations of various configurations included in the information processing apparatus 10 and executing various processes. Control. Details of processing executed in the control unit 11 will be described later.

  The communication unit 14 is a communication interface for communicating with various external devices.

  The storage unit 15 is configured by a storage device such as a hard disk. The storage unit 15 stores various types of information such as various programs necessary for execution of processing in the control unit 11 and information on processing results obtained by the control unit 11. Specific examples of information stored in the storage unit 15 will be described later.

  The display 16 is a display device that displays various types of information such as processing results by the control unit 11. The display 16 is configured by liquid crystal, organic EL, or the like.

  The operation unit 17 is a user interface for accepting operations on the information processing apparatus 10. The operation unit 17 includes an input device such as a keyboard and a mouse.

  With reference to FIG. 2, the functional configuration of the classifier construction apparatus 20 according to an embodiment will be described. The classifier construction apparatus 20 performs processing for constructing a classifier for classifying places where the user of the mobile terminal frequently visits. As the classification model, any model such as naive Bayes, k-nearest neighbor method, random forest, and the like can be used. Here, as a result of many experiments, the inventor of the present invention has found that a classifier constructed using a random forest is particularly effective in a classification model. However, the classification model in the present invention is not limited to a random forest.

  The classifier construction apparatus 20 includes a stay location detection unit 211, a clustering unit 212, a feature calculation unit 213, a classification determination unit 214, a classifier generation unit 215, and a database 217 as main functional configurations. The classifier construction apparatus 20 is implemented by the information processing apparatus 10, for example. The functions provided in the classifier construction device 20 are realized, for example, by the CPU 12 developing the program stored in the storage unit 15 or the like in the memory 13 and executing it in the control unit 11 of the information processing apparatus 10.

  The stay location detection unit 211 detects the stay location of the user of the mobile terminal (location where the user stayed for a predetermined time or more) using the positioning information of the mobile terminal stored in the database 217. The positioning information used here is information on the mobile terminal user's action base (for example, work place, home, etc. where the user goes frequently) in the database 217 out of the mobile terminal positioning information stored in the database 217. Is positioning information stored as user information. In other words, the stay detection unit 211 performs processing using the supervised positioning information in order to construct a classifier. In addition, although the stay location can be detected by an arbitrary method, in this embodiment, the stay location detection unit 211 first refers to the positioning information stored in the database 217 and determines the positioning position and the positioning time. A set of the positioning information satisfying the above relationship is determined, and a position determined based on the set is detected as a staying place. The predetermined relationship is, for example, a relationship in which the distance between positioning positions is within a predetermined distance (for example, within 200 m), and the time difference between positioning times is within a predetermined time (for example, within 10 minutes). . By detecting the place of stay in this way, a position where the user of the mobile terminal simply passes can be excluded from the place of stay. Moreover, as a position of a stay place, it can be set as the gravity center of the area | region surrounding the positioning position of all the positioning information contained in the set of positioning information, or the latitude and longitude of arbitrary positions in the said area, for example. Moreover, it can be set as the area | region which gave a certain amount of width without making the position of a stay place a point.

  FIG. 4A shows an example of the positioning information of the mobile terminal stored in the database 217. In this example, the positioning information includes information on the portable terminal ID, the positioning time, and the positioning position (latitude, longitude).

  FIG. 4B shows an example of user information stored in the database 217. The user information includes user ID, portable terminal ID, home area ID, work area ID, and other frequently used area ID information. The mobile terminal ID is identification information of a mobile terminal possessed by the corresponding user. The home area ID is identification information of an area where the corresponding user's home is located. The work area ID is identification information of an area where the corresponding user's work place is located. The other frequently-used area ID is identification information of other areas where the user frequently goes other than home and work place.

  The clustering unit 212 performs clustering on the stay place detected by the stay place detection unit 211 based on at least the position of the stay place. Specifically, the clustering unit 212 outputs an area where the stay places of the same mobile terminal are present at a predetermined density or more as a cluster of stay places of the user of the mobile terminal. The cluster output in this manner indicates a region where the user of the mobile terminal often stays as the number of staying places is increased. In this way, rather than simply counting the number of positioning points, it is possible to grasp the stay frequency closer to reality by grasping the stay frequency of the user of the mobile terminal by the number of stay places included in the cluster. .

  The feature calculation unit 213 calculates, for the clusters output from the clustering unit 212, feature values defined in advance for the stay destination cluster. The type of feature whose value is calculated is arbitrary. The feature calculation unit 213, for example, based on the information on the place of stay included in the cluster, stay rate, stay number rank, minimum stay time, maximum stay time, average stay time, stay days, daytime stay rate, and night stay rate Can be calculated as features. Although the feature to be calculated is not limited to these, each feature will be described below.

(1) Stay rate: The ratio of the number of stays included in the cluster (target cluster) that is the target of feature amount calculation to the number of stays included in all clusters output by the clustering unit 212 for the same mobile terminal It is.

(2) Stay number rank: The rank of clusters based on the number of stay places included in the target cluster in all clusters detected by the clustering unit 212 for the same mobile terminal. The greater the number of stays included in a cluster, the higher the stay rank of that cluster.

(3) Minimum stay time, maximum stay time, average stay time: These are the minimum value, maximum value, and average value of stay times for the stay places included in the target cluster. With these characteristics, for example, the tendency of whether the stay time in the cluster is long or short can be grasped.

(4) Stay days: The number of days staying in the target cluster. A day in which one or more mobile terminals appear on the same day in the cluster is counted as one day.

(5) Daytime stay rate, nighttime stay rate: ratio of time (or number of times) staying in the daytime (for example, 9:00 to 18:00) with respect to the total staying time (or number of times) in the target cluster, or nighttime in the cluster It is a ratio of the time (or the number of times) stayed at (for example, 22:00 to 6:00).

  The classification determination unit 214 refers to the position of the user's action base indicated in the user information of the mobile terminal stored in the database 217, and the cluster generated by the clustering unit 212 for the user is classified as the action base. It is determined whether or not. For example, the classification determination unit 214 uses the home area ID of the user information of the mobile terminal in which the position of the cluster generated by the clustering unit 212 for a certain user is stored in the database 217 for that user (see FIG. 4B). It is determined whether or not it corresponds to the position indicated by. Similarly, the classification determination unit 214 determines whether or not the positions of the clusters generated by the clustering unit 212 correspond to the positions indicated by the work area ID and other frequent area IDs in the user information.

  The classifier generation unit 215 performs supervised learning based on the feature value calculated by the feature calculation unit 213 and the determination result by the classification determination unit 214, uses the feature as an explanatory variable, and the stay destination cluster is a predetermined number. A classifier having the objective variable as to whether it is classified as an action base is generated. Further, the classifier generation unit 215 does not use all the features whose values are calculated by the feature calculation unit 213 to generate a classifier, but selects features that have an important influence on the cluster classification. It may be used. The selection of the feature that has an important influence on the classification of the cluster is performed using, for example, information gain. Further, the classifier is constructed based on an arbitrary learning algorithm. Information of the constructed classifier is stored in the database 217 as classifier information. The feature selected by the classifier generation unit 215 as a feature that has an important influence on the classification of the cluster greatly depends on information used for constructing the classifier. However, as a reference, as a result of repeated experiments, the inventors of the present invention have found that the above-described eight features often have an important influence as features used for cluster classification. Moreover, although these characteristics were described in order of (1) to (5), the characteristics are described in order from the one having the greatest influence.

  The database 217 stores various kinds of information such as information necessary for the above-described processing executed in the classifier construction apparatus 20 and information generated by the processing. The database 217 mainly stores, for example, positioning information, user information, classifier information, place of stay information, cluster information, and map information.

  Next, the functional configuration of the classification device 30 according to an embodiment will be described with reference to FIG. The classification device 30 uses the classifier constructed by the classifier construction device 20 to perform processing for classifying places (clusters of staying places) where the user of the mobile terminal frequently visits. The classification device 30 includes a stay location detection unit 221, a clustering unit 222, a feature calculation unit 223, a classification execution unit 224, and a database 227 as main functional configurations. The classification device 30 is implemented by the information processing device 10, for example. The functions provided in the classification device 30 are realized, for example, by the CPU 12 developing the program stored in the storage unit 15 or the like in the memory 13 and executing it in the control unit 11 of the information processing apparatus 10.

  The stay location detection unit 221 detects the stay location of the user of the mobile terminal (location where the user stayed for a predetermined time or more) using the positioning information of the mobile terminal stored in the database 227. The positioning information used here is positioning information that is a target of classification processing of places where the user goes frequently among the positioning information of the mobile terminal stored in the database 227. As the method for detecting the staying place, any method is used as described above.

  The clustering unit 222 performs clustering on the stay location detected by the stay location detection unit 221 based on at least the position of the stay location. Specifically, the clustering unit 222 outputs an area where the stay places of the same mobile terminal are present at a predetermined density or more as a cluster of the stay places of the user of the mobile terminal.

  The feature calculation unit 223 calculates, for the clusters output from the clustering unit 222, feature values defined in advance for the stay destination cluster. The type of feature whose value is calculated is arbitrary. The feature calculation unit 223, for example, based on the information on the place of stay included in the cluster, the stay rate, stay number rank, minimum stay time, maximum stay time, average stay time, stay days, daytime stay rate, and night stay rate Can be calculated as features. The meaning of each feature is as described above.

  The classification execution unit 224 uses the feature value calculated by the feature calculation unit 223 for the cluster output from the clustering unit 222, and uses the classifier generated by the classifier generation unit 215 and stored in the database 227. Output cluster classification results. For example, the classification execution unit 224 outputs a classification result indicating whether the cluster is classified into any action base including “residential area”, “work area”, and “other frequent areas”.

  Further, the classification execution unit 224 can correct the classification result with higher accuracy by using the map information stored in the database 227. For example, the classification result can be corrected using house map information included in the map information. A specific example will be described below. The residential map information includes information on whether the building is a residential house or a business building. The map is divided by a grid of a predetermined area (for example, 250 m square), and the number of residential houses and the number of business buildings in each grid are calculated. Here, according to a survey in a certain area by the inventor of the present invention, (A) the ratio of mobile terminal users who have a residential area in a grid where the number of residential houses is 50% or more is 72. It was 61%. Moreover, the ratio of the user of the portable terminal which has a work area in the grid with the ratio of (B) residential houses less than 50% was 74.31%. Furthermore, (C) a grid of 80.12%, there are more users who use the grid as a residential area than users who use the grid as a work area. When the numerical values of these (A), (B), and (C) (or any one of them) are compared with the aggregate value of the classification result by the classification execution unit 224, and there is a difference greater than a predetermined threshold , Correct the classification results so that the difference is small (for example, if the number of “residential areas” in the classification results is too large, the clusters classified as “residential areas” with a low likelihood are Make corrections to assign to the classification). That is, the classification execution unit 224 can correct the classification result by using the ratio of the number of residential buildings to the total number of buildings existing in the predetermined area.

  The database 227 stores various types of information such as information necessary for the above-described processing executed in the classification device 30 and information generated by the processing. The database 227 mainly stores, for example, positioning information, user information, classifier information, place of stay information, cluster information, and map information. In particular, the classifier information stored in the database 227 includes information on the classifier generated by the classifier generation unit 215.

  Next, the flow of processing executed in the classifier construction device 20 and the classification device 30 will be described with reference to FIGS. Each processing step included in the processing flow described below can be executed in any order or in parallel as long as there is no contradiction in the processing contents. These steps may be added. Further, a step described as one step for convenience can be executed by being divided into a plurality of steps, while a step described as being divided into a plurality of steps for convenience can be grasped as one step.

  FIG. 5 shows a flow of processing executed in the classifier construction apparatus 20 in order to construct a classifier for classifying places where the user of the mobile terminal frequently visits (cluster of staying places). Details of the processing in each step are as described above, and are omitted here.

  First, in step S <b> 11, the stay location detection unit 211 includes, as user information, information on the action base of the user of the mobile terminal (where the user frequently goes) in the database 217 among the positioning information of the mobile terminal stored in the database 217. Acquire or refer to stored positioning information. In step S <b> 12, the stay location detection unit 211 detects the stay location (location where the user stayed for a predetermined time or more) of the user of the mobile terminal using the acquired or referred positioning information.

  In step S <b> 13, the clustering unit 212 performs clustering on the stay location detected by the stay location detection unit 211 based on at least the position of the stay location. Specifically, the clustering unit 212 outputs a region where a predetermined number of places of stay for the same mobile terminal exist in a region of a predetermined area as a cluster of the user's stay location of the mobile terminal.

  In step S <b> 14, the feature calculation unit 213 calculates, for the clusters output from the clustering unit 212, feature values that are predefined for the stay destination cluster. The type of feature whose value is calculated is arbitrary as described above. The feature calculation unit 213, for example, based on the information on the place of stay included in the cluster, stay rate, stay number rank, minimum stay time, maximum stay time, average stay time, stay days, daytime stay rate, and night stay rate Can be calculated as features.

  In step S15, the classification determination unit 214 refers to the position of the user's action base indicated in the user information of the mobile terminal stored in the database 217, and the cluster generated by the clustering part 212 for the user is the action base. It is determined whether it is classified into. For example, the classification determination unit 214 determines whether the position of the cluster generated by the clustering unit 212 for a certain user corresponds to the position indicated by the home area ID of the user information of the mobile terminal stored in the database 217 for that user. Determine whether.

  In step S16, the classifier generation unit 215 uses the cluster feature value obtained as a result of the calculation in step S14 and the determination result in step S15 to select a feature that has an important influence on the cluster classification. To do. The selection of the feature that has an important influence on the classification of the cluster is performed using, for example, information gain.

  In step S17, the classifier generation unit 215 performs supervised learning based on the feature value for the feature selected in step S16 and the determination result by the classification determination unit 214 in step S15, and uses the feature as an explanatory variable. Then, a classifier having the objective variable as to whether or not the stay location cluster is classified into a predetermined action base is generated. The generation of the classifier is performed based on an arbitrary learning algorithm. The generated classifier information is stored in the database 217 as classifier information.

  As described above, according to the processing shown in FIG. 5, it is possible to construct a classifier for classifying places (clusters) where users often stay with higher accuracy.

  FIG. 6 shows a process executed by the classification device 30 to classify places where the user of the mobile terminal frequently uses the classifier constructed by the classifier construction device 20. Details of the processing in each step are as described above, and are omitted here.

  First, in step S21, the stay location detection unit 221 acquires or refers to positioning information that is a target of classification processing of places where the user goes frequently from the positioning information of the mobile terminal stored in the database 227, and in step S22 Detecting the place of stay of the user of the mobile terminal. As the method for detecting the staying place, any method is used as described above.

  In step S23, the clustering unit 222 performs clustering on the stay location detected by the stay location detection unit 221 based on at least the position of the stay location. Specifically, the clustering unit 222 outputs an area where the stay places of the same mobile terminal are present at a predetermined density or more as a cluster of the stay places of the user of the mobile terminal.

  In step S <b> 24, the feature calculation unit 223 calculates, for the clusters output from the clustering unit 222, feature values that are predefined for the stay destination cluster. The type of feature whose value is calculated is arbitrary as described above. The feature calculation unit 223, for example, based on the information on the place of stay included in the cluster, the stay rate, stay number rank, minimum stay time, maximum stay time, average stay time, stay days, daytime stay rate, and night stay rate Can be calculated as features.

  In step S <b> 25, the classification execution unit 224 uses the feature value calculated by the feature calculation unit 223 for the cluster output from the clustering unit 222 to generate the classification generated by the classifier generation unit 215 and stored in the database 227. The cluster classification result is output by the device.

  As described above, according to the processing shown in FIG. 6, the place where the user of the mobile terminal goes frequently is classified using the classifier constructed by the classifier construction device 20 and the positioning information of the mobile terminal. Classification can be performed with higher accuracy.

  The present invention is not limited to the above-described embodiment, and can be implemented in various other forms without departing from the gist of the present invention. The above-described embodiment is merely an example in all respects, and is not construed as limiting.

  DESCRIPTION OF SYMBOLS 10 Information processing apparatus, 11 Control part, 12 CPU, 13 Memory, 14 Communication part, 15 Storage part, 16 Display, 17 Operation part, 20 Classifier construction apparatus, 30 Classification apparatus

Claims (10)

Based on the positioning information of the mobile terminal, a generation unit that detects a stay location of the user of the mobile terminal and generates a cluster of the detected stay location (hereinafter referred to as a stay location cluster);
For the generated stay place cluster, a calculation means for calculating a value of a feature defined for the stay place cluster;
A determination unit that refers to position information of a predetermined action base associated with the user of the mobile terminal and determines whether the generated stay place cluster is classified into the predetermined action base;
A classifier that performs supervised learning based on the calculated feature value and the determination result, uses the feature as an explanatory variable, and sets whether the stay destination cluster is classified as a predetermined action base as an objective variable A classifier constructing device comprising: classifier generating means for generating The features are:
(A) The ratio of the number of staying places included in the target staying place cluster to the number of staying places included in all staying place clusters for the same mobile terminal,
(B) The rank of the target stay location cluster based on the number of stay locations included in the target stay location cluster in all the stay location clusters for the same mobile terminal,
(C) the minimum value of the stay time obtained for the stay places included in the target stay place cluster;
(D) the maximum value of the stay time obtained for the stay places included in the target stay place cluster;
(E) the average value of the stay time obtained for the stay places included in the target stay place cluster;
(F) Number of stays in the target stay cluster,
(G) the ratio of time spent in the day to the total stay time in the target stay cluster, and (h) the ratio of time spent in the night to the total stay time in the target stay cluster,
The classifier construction apparatus according to claim 1, comprising at least one of the following. The positioning information includes a positioning position and a positioning time,
The generating means is a detecting means for determining a set of positioning information satisfying a predetermined relationship between a positioning position and a positioning time, and detecting a position determined based on the set as a staying place, wherein the predetermined relationship is 3. The classifier construction apparatus according to claim 1, further comprising a detection unit in which a distance between the positioning positions is within a predetermined distance and a time difference in positioning time is within the predetermined time. Storage means for storing the classifier;
Based on the positioning information of the mobile terminal, a generation unit that detects a stay location of the user of the mobile terminal and generates a cluster of the detected stay location (hereinafter referred to as a stay location cluster);
For the generated stay place cluster, a calculation means for calculating a value of a feature defined for the stay place cluster;
Classifying means for classifying the stay destination clusters generated by the generating means by the stored classifier using the calculated feature value as an explanatory variable,
The classifier detects a staying place of the user of the mobile terminal based on the positioning information of the mobile terminal, generates a staying place cluster for the detected classifier, and the staying place cluster for the generated staying place cluster for the classifier The characteristic value defined for the mobile terminal is calculated, the location information of the predetermined action base associated with the user of the mobile terminal is referred to, and the generated classifier stay location cluster becomes the predetermined action base. It is determined whether or not it is classified, and supervised learning is performed based on the calculated feature value and the determination result, and the feature cluster is used as an explanatory variable, and the stay destination cluster is classified as a predetermined action base. A classifier that is a classifier generated using whether or not an objective variable is used.   The classification apparatus according to claim 4, wherein the classification unit corrects the classification result by using residential land map information.   The classification apparatus according to claim 5, wherein the modification includes modifying the classification result by using a ratio of the number of residential buildings to the total number of buildings existing in a predetermined area. A classifier construction method implemented in an information processing apparatus including a control unit,
The control unit detects a place of stay of the user of the mobile terminal based on the positioning information of the mobile terminal, and generates a cluster of the detected stay place (hereinafter referred to as a stay place cluster);
The control unit calculates, for the generated stay place cluster, a feature value defined for the stay place cluster; and
Determination that the control unit refers to position information of a predetermined action base associated with the user of the mobile terminal and determines whether the generated stay place cluster is classified as the predetermined action base Steps,
Whether the control unit performs supervised learning based on the calculated feature value and the determination result, uses the feature as an explanatory variable, and determines whether the stay destination cluster is classified as a predetermined action base And a classifier generation step for generating a classifier as a variable. A classification method implemented in an information processing apparatus including a control unit and storage means for storing a classifier,
The control unit detects a place of stay of the user of the mobile terminal based on the positioning information of the mobile terminal, and generates a cluster of the detected stay place (hereinafter referred to as a stay place cluster);
The control unit calculates, for the generated stay place cluster, a feature value defined for the stay place cluster; and
The control unit includes a classification step of classifying the stay destination cluster generated by the generation step by the stored classifier using the calculated feature value as an explanatory variable.
The classifier detects a staying place of the user of the mobile terminal based on the positioning information of the mobile terminal, generates a staying place cluster for the detected classifier, and the staying place cluster for the generated staying place cluster for the classifier The characteristic value defined for the mobile terminal is calculated, the location information of the predetermined action base associated with the user of the mobile terminal is referred to, and the generated classifier stay location cluster becomes the predetermined action base. It is determined whether or not it is classified, and supervised learning is performed based on the calculated feature value and the determination result, and the feature cluster is used as an explanatory variable, and the stay destination cluster is classified as a predetermined action base. A classification method, characterized by being a classifier generated using whether or not an objective variable is used. Computer
Based on the positioning information of the mobile terminal, a generating means for detecting a place of stay of the user of the mobile terminal and generating a cluster of the detected stay place (hereinafter referred to as a stay place cluster),
A calculation means for calculating a value of a feature defined for the stay place cluster for the generated stay place cluster;
Determination means for referring to position information of a predetermined action base associated with the user of the mobile terminal and determining whether the generated stay place cluster is classified as the predetermined action base;
A classifier that performs supervised learning based on the calculated feature value and the determination result, uses the feature as an explanatory variable, and sets whether the stay destination cluster is classified as a predetermined action base as an objective variable The program for functioning as a classifier construction apparatus provided with the classifier production | generation means which produces | generates. Computer
Storage means for storing a classifier;
Based on the positioning information of the mobile terminal, a generating means for detecting a place of stay of the user of the mobile terminal and generating a cluster of the detected stay place (hereinafter referred to as a stay place cluster),
A calculation means for calculating a value of a feature defined for the stay place cluster for the generated stay place cluster;
Classifying means for classifying the stay destination cluster generated by the generating means by the stored classifier using the calculated feature value as an explanatory variable,
The classifier detects a staying place of the user of the mobile terminal based on the positioning information of the mobile terminal, generates a staying place cluster for the detected classifier, and the staying place cluster for the generated staying place cluster for the classifier The characteristic value defined for the mobile terminal is calculated, the location information of the predetermined action base associated with the user of the mobile terminal is referred to, and the generated classifier stay location cluster becomes the predetermined action base. It is determined whether or not it is classified, and supervised learning is performed based on the calculated feature value and the determination result, and the feature cluster is used as an explanatory variable, and the stay destination cluster is classified as a predetermined action base. A program for functioning as a classification device, which is a classifier generated using whether or not an objective variable is used.

Images