WO2017071124A1

WO2017071124A1 - System and method for location and behavior information prediction

Info

Publication number: WO2017071124A1
Application number: PCT/CN2016/070989
Authority: WO
Inventors: 蒋昌俊; 闫春钢; 陈闳中; 丁志军; 徐兵
Original assignee: 同济大学
Priority date: 2015-10-28
Filing date: 2016-01-15
Publication date: 2017-05-04
Also published as: CN105357637B; CN105357637A

Abstract

A method for location and behavior information prediction, firstly acquiring the location and behavior information of the a mobile client and storing same, based on classification, in a data storage structure; traversing the data storage structure, circularly performing determination to obtain a cycle; and predicting, according to the cycle, customer location and behavior information. A user location and behavior information recording method for recording and sending the current location and behavior information, comprising: a location recording module collecting the current location and behavior information; the location recording module sending the collected current location and behavior information to an information transmitting module; and the information transmitting module externally sending the current location and behavior information. A mobile client for collecting and sending user location and behavior information. A location and behavior information prediction server for predicting location and behavior information. A location and behavior information prediction system, comprising a mobile client and a location and behavior information prediction server.

Description

Position and behavior information prediction system and method

Technical field

The present invention relates to an information prediction system, and more particularly to a position and behavior information prediction system and method.

Background technique

In the field of LBSN (Location-based Social Network), the acquisition of user behavior patterns has attracted the attention of scholars and produced more research results. Among them, Zheng Yu and other researchers based on the user's driving trajectory, by mining the user's frequent movement sequence and establishing a probability transfer model, can better predict the user's next possible arrival location. In general, the user-based behavior pattern can effectively improve the accuracy of predicting the user's future arrival location. If the user's behavior pattern is combined with the recommendation system, the user can provide a recommendation service that is more in line with their living habits.

Periodic behavior is a kind of user behavior pattern. By mining the user's periodic behavior, it is possible to effectively discover the regularity of a user's behavior in time. At present, the main research directions of periodic behavior are divided into time series cycle acquisition and periodic mode acquisition. The former obtains the potential period in the time series through the relevant algorithm. The classical algorithm acquired in the time series period is the periodogram and the autocorrelation. When the periodogram is used alone, the period is inaccurate due to the spectral leakage problem, and it is used alone. At the time of autocorrelation, there are other timestamp events, which lead to the correct cycle is not obvious. Some researchers use the combination of cycle diagram and autocorrelation to effectively avoid the spectral leakage caused by the use of the periodic diagram alone. The cycle gets inaccurate.

In summary, the current time series period acquisition algorithm has mutual interference between periods when acquiring time series and multiple periods, resulting in inaccurate final results. And mining the user's periodic behavior needs to accurately obtain the user's change behavior at which stage of the current cycle, and the autocorrelation and cycle diagram can not mine this type of information.

Summary of the invention

In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a position and behavior information prediction system and method for solving the cycle of low spectral acquisition accuracy, periodic spectrum spectral leakage and autocorrelation in the prior art. Not obvious and other issues.

To achieve the above and other related objects, the present invention provides a location and behavior information prediction system and method.

Preferably, the location and behavior information prediction method is used to predict the location of the user's future arrival by mining the period of the user behavior; the location and behavior information prediction method mainly includes the following steps:

The first step is to obtain the location and behavior information of the mobile client separately;

The second step is to classify the acquired location and behavior information to form a sequence of location information and a sequence of behavior information, respectively;

In the third step, the position and behavior information of different classes are converted and processed and placed in the data storage structure, and the elements in each data storage structure can be randomly accessed; the data storage structure is traversed, the judgment conditions are cyclically executed, and multiple selections are performed. a different cycle;

The fourth step is to predict the location and behavior information of the mobile client according to the obtained period.

Preferably, the first step further comprises: determining a specific time period; and selecting location and behavior information in the specific time period from the received location and behavior information.

Preferably, the second step further comprises: classifying the location and behavior information in the specific time period; adding the category information generated in the classification process to each location and behavior information, and generating different types of location and behavior information according to the chronological order sequence.

Preferably, the third step further comprises: setting a fixed relationship between the location in the data storage structure and the location element; determining the element value according to the fixed relationship, and randomly traversing each element in the data storage structure; determining according to the setting Relationships filter out accurate cycles and save.

Preferably, a user location and behavior information recording method is used for recording current location and behavior information of the user and transmitting current location and behavior information to other receiving ends; and the method comprises the following steps:

The first step: the location record module collects current location and behavior information;

Step 2: The location recording module transmits the collected current location and behavior information to the information sending module;

The third step: the information sending module sends the current location and behavior information to the outside.

Preferably, the location recording module collects the current location and behavior information of the user, and transmits the collected current location and behavior information to the information sending module.

Preferably, a mobile client, comprising:

A location recording module that collects the current location and behavior information of the user.

The information sending module communicates with the location recording module; and is configured to send the current location and behavior information collected by the location recording module.

Preferably, a location and behavior information prediction server includes: a location and behavior information processing module, a multi-cycle acquisition module, and a prediction module;

Preferably, the location and behavior information processing module is configured to receive and store the current location and behavior information of the user, and obtain a set of specific time slot locations and behavior information of the specific mobile client from all current location and behavior information to form time series information. ;

Preferably, the multi-cycle acquisition module is in communication with the location and behavior information processing module for finding all periods in the time series information;

Preferably, the prediction module is in communication with the multi-cycle acquisition module for predicting location and behavior information of the user at a future point in time.

Preferably, the location and behavior information processing module comprises

An information storage module, configured to store current location and behavior information sent by the mobile client;

An information acquisition module, the module is in communication with the information storage module, and is configured to obtain current location and behavior information;

a location classification module, the module is in communication with the information acquisition module, and is configured to obtain current location and behavior information;

A timing module that communicates with the location classification module for generating a sequence of formation location and behavior information.

Preferably, the multi-cycle acquisition module communicates with the location and behavior information processing module to receive the location and behavior information sequence; the multi-cycle acquisition module is configured to determine a plurality of cycles.

Preferably, the prediction module communicates with the multi-cycle acquisition module to predict the location of the mobile client according to the period.

Preferably, a location and behavior information prediction system includes a mobile client and a location and behavior information prediction server;

Preferably, there are multiple mobile clients, and each mobile client continuously collects current location and behavior information of the user.

Preferably, the location and behavior information prediction server is configured to predict the location and behavior information of the mobile client at a future time point from different time points.

As described above, the position and behavior information prediction system and method provided by the present invention have the following beneficial effects:

The present invention provides a position and behavior information prediction system and method, and the present invention can effectively avoid adverse effects such as cycle length differences and random fluctuations as compared with the conventional use cycle diagram and autocorrelation and the combination of the two. Obtaining the influence of cycle precision, and reducing the mutual interference between different cycles in the multi-cycle acquisition process, can more accurately acquire the multi-periodic behavior in the time series, and solve the problem that the cycle obtained in the traditional method is not obvious. By mining the user's periodic behavior, it is possible to effectively discover the regularity of a user's behavior in time, and can combine with the recommendation system to produce technical benefits. The position and behavior information prediction method provided by the invention does not need to search for a potential period or a time series period, but through random traversal and judgment of the data storage structure, and filters the period information with higher precision, thereby reducing the period of the spectrum acquisition process due to the spectrum leakage problem. The adverse effects. The invention can accurately obtain the user's behavior in the current cycle by traversing and judging the data structure, and determine the judgment condition of each step traversal process by using parameters such as the number of times of repeated occurrences, thereby excavating the traditional autocorrelation and The exact period information that the periodogram cannot mine. The invention can simultaneously acquire the occurrence time of the periodic behavior, and solves the problem that the periodic graph and the autocorrelation cannot predict the periodic behavior of the user. Point problem.

DRAWINGS

1 is a schematic diagram of a position and behavior information prediction system of the present invention;

2 is a schematic diagram showing a method for predicting location and behavior information according to the present invention;

3 is a schematic diagram of a location and behavior information processing module of the present invention;

Figure 4 is a schematic view showing the multi-cycle searching step of the present invention.

detailed description

The embodiments of the present invention are described below by way of specific embodiments, and those skilled in the art can readily understand other advantages and functions of the present invention from the disclosure.

Please refer to Figure 1 to Figure 4. It is to be understood that the structures of the present invention are intended to be in accordance with the disclosure of the specification, and are not intended to limit the scope of the invention. The technical significance, the modification of any structure, the change of the proportional relationship or the adjustment of the size, without affecting the effects and the achievable purposes of the present invention, should still fall within the scope of the present invention. Within the scope of the technical content. In the meantime, the terms "upper", "lower", "left", "right", "intermediate" and "one" as used in this specification are also for convenience of description, and are not intended to limit the present. The scope of the invention can be implemented, and the change or adjustment of the relative relationship is considered to be within the scope of the invention.

The first embodiment provides a location and behavior information prediction method, and provides a location and behavior information prediction server end;

As shown in FIG. 2, the location and behavior information prediction method is used to predict the location where the user arrives in the future by mining the period of the user behavior; the location and behavior information prediction method mainly includes the following steps:

The first step: obtaining the location and behavior information of the mobile client carried by the user;

The second step: classifying the location and behavior information to form a sequence of location and behavior information;

The third step: put different types of location and behavior information in different collections, the elements in each collection can be randomly accessed; dynamically allocate storage space, constantly update the location and behavior information in the collection; perform judgment conditions, evaluate each The periodic strength between the location and behavior information, obtaining multiple different periods;

Step 4: Predict the location where the mobile client arrives at a future point in time based on the different cycles obtained.

The first step further includes: after taking the location and behavior information in the database according to the main code, selecting the location in the specific time period and Behavioral information.

The second step further includes: classifying the location and behavior information in the specific time period; adding the category information generated in the classification process to each location and behavior information to obtain different types of location and behavior information sequences.

The third step further includes: setting a fixed relationship between the position in the data storage structure and the position element in the position and behavior information sequence; and obtaining the position and behavior information in the position and behavior information sequence by the fixed relationship The corresponding elements are placed in each position in the data storage structure in a bit position; a fixed relationship between the element values and the cycle reliability is set, and each element in the data storage structure is traversed to determine a period with high reliability and saved.

The principle of traversing the data storage structure in the third step is that there is a set of all the different time intervals in which all the specific class events occur, and the periods in the time series exist in the set. In this embodiment, the data storage structure selection matrix, because the matrix can better preserve such information, can initially remove the error period by creating a suspect periodic matrix, and can effectively improve the random access based on the characteristics of the two-dimensional array random access. The efficiency of accessing matrix elements. Based on the characteristics of time series period and the basic idea of matrix storage, a dynamic update matrix multi-cycle acquisition algorithm is constructed to further judge whether the support degree of the suspect period in the matrix is greater than the set support threshold. If the current suspect period is a real period, Then dynamically modify the elements in the matrix to reduce the number of algorithm judgments, improve the execution efficiency of the algorithm, and finally obtain all the periodic behaviors in the time series.

According to the actual situation in which the periodic behavior always occurs at the same time interval, and the characteristics of the matrix random access, a matrix as shown below is first established, wherein the first row and the first column are time stamps of the occurrence of the behavior. The elements in the matrix are the difference between the column timestamp and the row timestamp.

The matrix is a matrix that is actually processed by multi-periodic behavior, and includes true periodic behavior and non-authentic periodic behavior. The true periodic behavior has a support degree greater than the support threshold set by the algorithm, such as Equation 1.

Formula 1

Where reality_happened_time(x) represents the actual number of occurrences of the cycle, and theory_happened_time(x) represents the number of theoretical occurrences. In the matrix, the line number of each element is the timestamp of the first occurrence, and the value of the element is the period of the periodic behavior. Since the periodic behavior occurs in a fixed period, the timestamp of the theoretical occurrence can be obtained from the start time and the size of the period, and the number of occurrences is counted. When the theory occurs, the timestamps are sorted from small to large. If the first two consecutive timestamps are the first row and the second column is hit in the matrix, it is recorded as the actual occurrence. The total number of hits is the actual number of occurrences. , corresponding to the reality_happened_time(x) in Equation 1.

As shown in FIG. 4, the time series multi-cycle acquisition process in the present invention mainly includes the following steps:

The first step: the method uses the gradual increase of the matrix row number to simulate the same kind (the relationship between the timestamp category and the row or column) The sequence in which the states occur in the time series;

Step 2: Simulate the time interval between any timestamps by row corresponding subtraction;

The third step: in the suspected periodic matrix, after obtaining the set containing all the time intervals, the traversal of the entire matrix is started, and the line numbers of the elements larger than zero are stored as the suspected period;

The fourth step: obtaining a time stamp according to the currently obtained suspected cycle, and obtaining the number of theoretical occurrences;

Step 5: Sort the timestamps of the theory from large to small. If one of the successive timestamps is the first row and the second column is hit in the matrix, it is recorded as the first actual occurrence. The number of hits is the actual number of occurrences;

The sixth step: the actual number of occurrences (statistical acquisition) is more than the theoretical number, the actual occurrence of support is obtained, and compared with the threshold, greater than true, less than the cycle and then traversed;

Step 7: After the true storage, the zero is returned to the matrix, and after the matrix is dynamically updated, the updated matrix is looped through;

Step 8: Output all the filtered element values as the exact period size and the element line number as the first occurrence time;

As shown in FIG. 1, the second embodiment provides a user location and behavior information recording method for recording a current location and behavior information of a user; and transmitting current location and behavior information to other receiving ends;

The user location recording method includes the following steps:

The third step: the information sending module sends the current location and behavior information to the receiving end.

The location recording module continuously collects the current location and behavior information of the user at a fixed time interval, and transmits the collected current location and behavior information to the information sending module.

The third embodiment provides a mobile client, each mobile client having a unique number for the information sending module to send the current location and behavior information according to the unique number of the mobile client.

As shown in FIG. 1 , the mobile client includes at least: a location recording module and an information sending module;

The location recording module collects the current location and behavior information of the user, and transmits the current location and behavior information to the information sending module; the information sending module is configured to send the current location and behavior information outward.

As shown in FIG. 1 , a fourth embodiment provides a location and behavior information processing server, including at least: a location and behavior information processing module, a multi-cycle acquisition module, and a prediction module;

The location and behavior information processing module is configured to store the current location and behavior information of the user by each mobile client number; and obtain a specific time slot location and behavior information collection of the specific mobile client from all current location and behavior information, and Sorting the elements in the collection to form time series information;

The multi-cycle acquisition module is connected to the location and behavior information processing module for finding all periodic letters in the time series information interest;

The prediction module is coupled to the multi-cycle acquisition module for processing each cycle information to predict location and behavior information of the user at a future point in time.

As shown in FIG. 3, the location and behavior information processing module includes an information processing module, an information acquisition module, a location classification module, and a timing module; the information storage module stores the location and behavior information of each mobile client as a group according to the number; The obtaining module is connected to the information storage module, and obtains specific time slot location and behavior information of any mobile client from the information storage module; the location classification module is connected with the information acquiring module to classify the specific time period position and behavior information. The timing module is connected to the location classification module to form a sequence of positional and behavioral information in chronological order in the same location and behavioral information in a fixed time unit.

The multi-cycle acquisition module is coupled to the location and behavior information processing module for receiving the location and behavior information sequence formed by the location and behavior information processing module; the multi-cycle acquisition module finds each location and behavior information sequence from each location and behavior information sequence. The position and behavior information of the cycle in which the cycle is reproduced determines the cycle information.

The prediction module is connected to the multi-cycle acquisition module, and predicts various types of locations that the mobile client arrives at a future time point from different time points according to the period information.

A fifth embodiment provides a location and behavior information prediction system, including at least: a mobile client and a location and behavior information prediction server;

The mobile client is easy to carry and is used to collect the current location and behavior information of the mobile client itself;

The mobile client is configured to send the collected current location and behavior information to the location and behavior information prediction server;

The location and behavior information prediction server is configured to receive current location and behavior information of the user sent by the mobile client, and confirm various behavior activity periods of the user according to the current location and behavior information; location and behavior information prediction server The end is used to predict the location and behavior of the user at a future point in time based on the current location and behavior information.

The number of mobile clients is no less than two; each mobile client collects the user's current location and behavior information at regular intervals.

The location and behavior information prediction server is used to predict the location and behavior information of future time points of multiple mobile clients according to different periods from different time points.

In the fifth embodiment, in the location and behavior information prediction system and method (GPS coordinates and) provided by the present invention, the GPS acquisition module of the user mobile client is started to run as a background process (generally, the boot is set to boot), for each The user's current GPS latitude and longitude coordinates are collected in 5 seconds. The user uses the mobile phone number as the unique identifier (the main code in the database), and sends the latitude and longitude coordinates to the position and behavior information to predict the GPS coordinates in the location and behavior information processing module in the server. Storage module. (GPS)

The GPS coordinate storage module accepts the latitude and longitude coordinates sent by the client and saves it to the MySQL database.

The stay location acquisition module in the startup location and behavior information processing module acquires a stay point for each user, saves the entry time and departure time of the stay point, and saves the acquired stay point to the stay point table. The user's day of stay is usually obtained in a day-level task. Activate the stay point clustering module after completing the day of stay.

The stay point clustering module reads the data of the stay points of each user for the past 8 weeks (including the stay points of the day), and clusters the stay points using the OPTICS clustering algorithm. After the clustering is completed, the category information is added to the stay points. . Activate the cycle acquisition module.

The periodic acquisition module sequentially reads the stay point of a certain category of a certain user, and simultaneously generates a time series of the user reaching the stay point by using the time granularity of the time, and uses the time series multi-cycle acquisition algorithm to obtain the period of the user reaching the stay point. , save the cycle information to the database. Activate the user periodic behavior prediction module.

The periodic behavior of each user is read in turn, and the next time point at which the user arrives at the stay point of the category is predicted based on the starting time point, the period size, and the staying point category.

In the sixth embodiment, in the location and behavior information prediction system and method provided by the present invention, the behavior collection module of the user mobile client is started to run as a background process (generally set to start-up), and the current user is collected every 5 seconds. Behavior information, establish a behavioral causal connectivity graph, determine the behavior node of the behavior information in the causal connectivity graph, the user uses the mobile phone number as the unique identifier, and sends the behavior node information to the behavior information processing module in the behavior information prediction server. Node storage module.

The node storage module accepts the behavior node information sent by the client and saves it to the MySQL database.

The node acquisition module in the behavior information processing module is configured to acquire a behavior node for each user, save the start time and end time of the behavior node, and save the acquired behavior node to the node table. The behavior node of the user's day is usually obtained by a day-level task. The node clustering module is activated after all behavior nodes are acquired on the day of completion.

The node clustering module reads the behavior node data of each user in the past 8 weeks (including the behavior nodes of the day), and clusters the behavior nodes using the OPTICS clustering algorithm, and adds the category information to the behavior nodes after the clustering is completed. Activate the cycle acquisition module.

The periodic acquisition module sequentially reads the behavior node of a certain category of a certain user, and generates a time series of the behavior node of the user by using the time series multi-cycle acquisition algorithm to acquire the period of the behavior node, and saves the period information. To the database. Activate the user periodic behavior prediction module.

The periodic behavior of each user is read in turn, and the next time point of the behavioral node in the causal connected graph is predicted based on the starting time point, the period size, and the behavior node category, that is, the behavior of the user is predicted.

In the prior art, the main research directions of periodic behavior are divided into time series period acquisition and periodic pattern acquisition. The former obtains the potential period in the time series through the relevant algorithm. The classical algorithm acquired in the time series period is the periodogram and the autocorrelation. When the periodogram is used alone, the period is inaccurate due to the spectral leakage problem, and it is used alone. At the time of autocorrelation, there are other timestamp events, which lead to the correct cycle is not obvious. Some researchers use the combination of cycle diagram and autocorrelation to effectively avoid the spectral leakage caused by the use of the periodic diagram alone. The cycle gets inaccurate.

In summary, the current time series period acquisition algorithm has mutual interference between periods when acquiring time series and multiple periods, resulting in inaccurate final results. And mining the user's periodic behavior needs to accurately obtain the user's change behavior at which stage of the current cycle, and the autocorrelation and cycle diagram can not mine this type of information. The position and behavior information prediction system and method provided by the invention can effectively avoid the adverse effects such as the difference of the cycle length and the random fluctuation of the acquisition cycle precision, compared with the traditional use cycle diagram and autocorrelation and the combination of the two. The impact, while reducing the mutual interference between different periods in the multi-cycle acquisition process, can more accurately obtain the multi-periodic behavior in the time series, and solve the problem that the period obtained in the traditional method is not obvious; The periodic behavior can effectively discover the regularity of the user's behavior in time, and can be combined with the recommendation system to produce technical benefits; the invention does not need to find the potential cycle or time series cycle, but through the random traversal of the data storage structure. And judging, filtering the cycle information with higher precision, reducing the adverse effect of the spectral leakage problem on the cycle acquisition process; the invention can simultaneously acquire the time point of the periodic behavior, and solve the cycle diagram and the autocorrelation cannot predict the user periodicity The problem at the point in time when the behavior occurs; the present invention is capable of passing The traversal and judgment of the data structure accurately obtains which phase of the current cycle the user changes behavior, and uses the parameters such as the number of times of repeated occurrences to determine the judgment condition of each step of the traversal process, thereby excavating the traditional autocorrelation and the cycle diagram cannot be excavated. Precise cycle information.

The above embodiments are merely illustrative of the principles of the present invention and its effects, and are not intended to limit the present invention. Modifications or variations of the above-described embodiments may be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, all equivalent modifications or changes made by those skilled in the art without departing from the spirit and scope of the invention are still covered by the appended claims.

Claims

A method for predicting location and behavior information, comprising: the following steps:

The first step is to separately obtain the location and behavior information of a mobile client;

a second step of classifying the acquired location and behavior information to form a sequence of location information and a sequence of behavior information;

In the third step, the location and behavior information of different classes are converted and processed and placed in the data storage structure, and the elements in each of the data storage structures are randomly accessible; the data storage structure is traversed, and the judgment conditions are cyclically executed. Select multiple different cycles;

The fourth step is to predict the location and behavior information of the mobile client according to the obtained period.
The location and behavior information prediction method according to claim 2, wherein the first step further comprises: determining a specific time period; and selecting location and behavior information in the specific time period from the received location and behavior information.
The position and behavior information prediction method according to claim 2, wherein the second step further comprises: classifying the position and behavior information in the specific time period; and adding the category information generated in the classification process to each piece. In the position and behavior information, different types of position and behavior information sequences are generated in chronological order.
The position and behavior information prediction method according to claim 3, wherein the third step further comprises: a fixed relationship between the position in the set data storage structure and the position element; and obtaining the fixed relationship according to the fixed relationship The element value randomly traverses each element in the data storage structure; selects an accurate period according to the set judgment relationship and saves it.
A user location and behavior information recording method for recording a current location and behavior information of a user and transmitting the current location and behavior information to other receiving ends; and the method includes the following steps:

a first step: the location record module collects the current location and behavior information;

The second step: the location recording module transmits the collected current location and behavior information to the information sending module;

The third step: the information sending module sends the current location and behavior information to the outside.
The user location and behavior information recording method according to claim 5, wherein the location recording module collects current location and behavior information of the user, and transmits the collected current location and behavior information to the information sending module.
A mobile client, comprising:

A location recording module that collects the current location and behavior information of the user.

The information sending module is configured to communicate with the location recording module, and is configured to send the current location and behavior information collected by the location recording module.
A location and behavior information prediction server is characterized in that it comprises: a location and behavior information processing module, a multi-cycle acquisition module and a prediction module;

a location and behavior information processing module, configured to receive and store a current location and behavior information of the user, and obtain a set of specific time slot locations and behavior information of the specific mobile client from all current location and behavior information to form time series information;

a multi-cycle acquisition module, configured to communicate with the location and behavior information processing module, to find all cycles in the time series information;

The prediction module is in communication with the multi-cycle acquisition module for predicting location and behavior information of the user at a future point in time.
The location and behavior information prediction server according to claim 8, wherein said location and behavior information processing module comprises

An information storage module, configured to store current location and behavior information sent by the mobile client;

An information acquiring module, configured to communicate with the information storage module, to obtain the current location and behavior information;

a location classification module, configured to communicate with the information acquisition module, to classify the current location and behavior information;

The timing module is in communication with the location classification module for generating a location and behavior information sequence.
The location and behavior information prediction server according to claim 9, wherein said multi-cycle acquisition module communicates with said location and behavior information processing module to receive said location and behavior information sequence; said multi-cycle acquisition The module is used to determine multiple cycles.
The location and behavior information prediction server according to claim 10, wherein the prediction module is in communication with the multi-cycle acquisition module, and predicts a location of the mobile client according to the period.
A location and behavior information prediction system, comprising: the mobile client as claimed in claim 7 And the location and behavior information prediction server end according to any one of claims 8 to 11;
The location and behavior information prediction system according to claim 12, wherein the plurality of mobile clients are multiple, and each of the mobile clients continuously collects current location and behavior information of the user.
The location and behavior information prediction system according to claim 13, wherein the location and behavior information predicts location and behavior information of the mobile client for future time points from different time points. Make predictions.