WO2012034388A1

WO2012034388A1 - Method and apparatus for user behaviors statistics based on user events

Info

Publication number: WO2012034388A1
Application number: PCT/CN2011/071904
Authority: WO
Inventors: 马兆勉
Original assignee: 中兴通讯股份有限公司
Priority date: 2010-09-13
Filing date: 2011-03-17
Publication date: 2012-03-22
Also published as: CN101951623B; CN101951623A

Abstract

A method for user behaviors statistics based on user events is disclosed in the present invention, which includes: acquiring dimensionality information of current state of a user from user basic information after receiving a user event message; accumulating counts value of event relative indexes in a corresponding statistical object example according to the dimensionality information and event types of the current state of the user (102). An apparatus for user behaviors statistics based on user events is simultaneously disclosed in the present invention. The method and apparatus of the present invention enable to compress the user behaviors according to the dimensionality space, enable to save system resources, also enable to flexibly analyze and count user behaviors in any dimensionality or combined dimensionalities based on the statistical results of user behaviors according to the user dimensionalities, and also enable to automatically remove the invalid dimensionality space event.

Description

User behavior statistics method and device based on user event

The invention relates to statistical technology in the field of communication, in particular to a user behavior statistical method and device based on user events. Background technique

With the development of the mobile network, the operator's concern has changed from the bearer profit model to the content service profit model. Therefore, the traditional performance statistics object can no longer meet the operator's requirements for the refined operation of the mobile service. The analysis of user behavior has become the focus of operators' attention and the basis for improving profitability. User behavior analysis can be performed by counting the user's call records and media message content, and these call records and media messages contain far more content than traditional performance statistics objects, based on call records and media messages. Statistics and analysis enable in-depth analysis of a range of metrics such as system performance and user behavior to obtain more valuable information.

For a mobile user, his business behavior has multiple attributes, such as: The user location must belong to a certain location area, routing area, service area, etc., and the access method must be attributed to some access method, such as the second generation. Mobile communication technology (2G, Second Generation), third-generation mobile communication technology (3G, 3rd-Generation), or fourth-generation mobile communication technology (4G, Fourth Generation), whose access point must belong to an access point name (APN, Access Point Name) „ In the following description, attributes such as user location, access method, and home APN related to the event are called dimensions, and the type of business behavior of the user is called an event, and the specific business process is called. The statistical results are called indicators.

In the analysis of user behavior, operators need to be able to analyze user behavior from multiple dimensions or combined dimensions and multiple metrics, for example, for general packet wireless service technology. Serving GPRS Support Node (SGSN), the dimension that the operator pays attention to, may be routing area, location area, service area, tracking area, access point, access mode, service type, user category, and terminal type. A specific or combined dimension in the series dimension, the concern indicators include the number of attached users, the number of activated users, the attach success rate, the activation success rate, the paging success rate, the delay, the quality of service (QoS), and the terminal type. A series of statistical results such as distribution, therefore, the statistical system is required to freely combine statistical results to achieve flexible query.

For the statistical analysis needs of this multi-dimensional, multi-index flexible operation, the usual processing method is: Establish a Cartesian model of the attention indicator in the statistical system. Specifically, assume that there are M dimensions in the user information, and the indicators that the operator pays attention to. Involving N dimensions, where N≤M, D ₁ D ₂ D _n represents N dimensions, and the corresponding number of dimensions is N, N ₂

N _n , then the total number of records for the Cartesian model, can be expressed as:

N=NixN ₂ x xN _n

However, for the field of mobile communications, the number of user-related dimensions is usually large, for example: SGSN dimensions include routing area, location area, access point, and service area. If the above formula is used, the total number of records corresponding to it will be very large. According to the method of allocating memory and performing arithmetic operations, the resources and performance of the existing system are often difficult to meet the requirements. Therefore, the above method needs to be improved. The following two solutions are available in the prior art:

First, for the above model, the dimensional space is compressed according to the relationship of the dimensions. For example, a fixed location area in the network model must belong to a specific routing area. For other non-associated routing areas, it is impossible for the location area to be combined with other non-associated routing areas. In this way, some combinations of dimensions that are unlikely to occur can be eliminated, thereby achieving compression of the dimensional space. This method can save some memory resources. However, the dimensional compression method is too complicated. If the network model of the network element changes, it is difficult to implement automatic processing.

Second, according to the predefined model, a fixed statistical model is established for the dimensions and indicators of the operator's attention, and statistical analysis is performed separately. This statistical method is more suitable for real-time monitoring, In order to obtain a relatively high compression ratio, however, since the method is based on a fixed model and a fixed index, the statistical function is single, and it is difficult to meet the flexible combination requirements of the dimensions.

In summary, the existing user behavior statistics method cannot implement multi-dimensional statistics and combined multi-dimensional statistics of event-based user behavior. Summary of the invention

In view of this, the main object of the present invention is to provide a user behavior statistical method and apparatus based on user events, which can implement multi-dimensional statistics and combined multi-dimensional statistics of event-based user behavior.

In order to achieve the above object, the technical solution of the present invention is achieved as follows:

The present invention provides a user behavior statistics method based on user events, the method includes: after receiving a user event message, obtaining dimension information of a current state of the user from the currently saved basic user information;

The count value of the event-related indicator in the corresponding statistical object instance is accumulated according to the dimension information and the event type of the current state of the user.

In the foregoing solution, before obtaining the dimension information of the current state of the user from the currently saved user basic information, the method further includes:

When the user event packet carries the dimension in which the event changes, the saved user basic information is updated according to the user dimension carried in the user event packet.

In the above solution, the sum of the event-related indicators in the corresponding statistical object instance is accumulated according to the dimension information and the event type of the current state of the user, and is:

According to the dimension information of the currently saved user state, the corresponding statistical object instance is searched. If the corresponding statistical object instance is found, the count value of the event related indicator in the statistical object instance is incremented by 1 according to the event type, if no corresponding corresponding is found. For the statistical object instance, a new statistical object instance is created, and the count value of the event-related indicator in the statistical object instance is incremented by one according to the event type. In the foregoing solution, the searching for the corresponding statistical object instance according to the dimension information of the currently saved user state is:

The index corresponding to the dimension information is calculated according to the dimension information of the currently saved user state; and the statistical object instance corresponding to the index is searched according to the index corresponding to the dimension information.

In the above solution, the method further includes:

After the timer expires, save the statistics and clear all statistical object instances to restart statistics.

In the above solution, the method further includes:

After the original statistics are completed, the statistical results are further analyzed and calculated according to the user's requirements, and the operation results are output.

The present invention also provides a user behavior statistics device based on a user event, the device comprising: a message analysis module and a statistics module;

The packet analysis module is configured to: after receiving a user event packet, obtain the dimension information of the current state of the user from the currently saved user basic information, and send the obtained current dimension information of the user to the statistics module;

The statistics module is configured to: after receiving the dimension information of the current state of the user sent by the ^ text analysis module, accumulate the count value of the event-related indicator in the corresponding statistical object instance according to the dimension information and the event type of the current state of the user.

In the above solution, the device further includes a storage module, configured to save user basic information and a statistical object instance, where the user basic information includes dimension information of a current state of the user.

In the foregoing solution, the device further includes a collecting module, configured to receive a user event packet, and when the user event packet carries the dimension in which the event changes, the storage module is updated according to the dimension carried in the user event packet. Saved user basic information.

In the above solution, the device further includes: a timer and a clearing module; wherein, the timer is configured to trigger the clearing module after the timeout; The clearing module is used to save the statistics after being triggered by the timer, and clear all statistical object instances.

In the above solution, the device further includes a calculation module, configured to further analyze and calculate the statistical result according to the user requirement after the statistics are completed, and output the operation result.

The user event-based user behavior statistics method and device provided by the present invention, after receiving the user event message, accumulate the count value of the event-related indicator in the corresponding statistical object instance according to the dimension information and the event type of the current state of the user. In this way, user behavior can be classified and compressed according to user dimension attributes and event types, that is, statistics of user behavior multi-dimensional and combined dimensions can be realized, thereby saving system resources; in addition, for example, in a mobile communication system, The specific location area must belong to a specific routing area, and the combination of the location area and the routing area without the affiliation relationship will inevitably result in no user events. Therefore, the solution of the present invention can automatically implement the culling of invalid dimensional space events. .

The index corresponding to the dimension information is calculated according to the current dimension information of the user state. The statistical object instance corresponding to the index is searched according to the index corresponding to the dimension information, so that the statistical object instance corresponding to the dimension information can be quickly found.

In addition, after the original statistics are completed, according to the user's needs, the user analyzes the object, further analyzes and calculates the statistical result, and outputs the operation result for the user, so that the second operation can be realized on the basis of the original statistical result. Further meet user needs. DRAWINGS

1 is a schematic flowchart of a user behavior statistics method based on user events according to the present invention; FIG. 2 is a schematic diagram of original statistical results of Embodiment 1;

3 is a schematic diagram of a result of a dimension combination query of Embodiment 1;

4 is a schematic diagram showing an output result of the activation success rate of the first embodiment;

FIG. 5 is a schematic structural diagram of a user behavior statistics device based on user events according to the present invention. detailed description

The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

The event statistics method based on the user behavior of the present invention, as shown in FIG. 1 , includes the following steps: Step 101: After receiving a user event packet, obtain the dimension information of the current state of the user from the currently saved basic user information;

Here, the user event message includes a user identifier, an event type, and an execution result, and the currently saved user basic information can be found according to the user identifier, that is, the user identifier is index information of the currently saved user basic information; The user basic information includes the dimension information of the current state of the user; the dimension information of the current state of the user refers to specific information of all dimensions of the current state of the user; after receiving the user event packet, the current identifier of the user is used to find the current user identifier according to the user event packet. The saved user basic information, if not found, indicates that the user event message of the user is received for the first time, the basic information of the user is saved, and after receiving the user event message again, the basic information of the user is first. Update based on the second save;

Before obtaining the dimension information of the current state of the user from the currently saved user basic information, the method may further include:

When the user event packet carries the dimension in which the event changes, the saved user basic information is updated according to the dimension carried in the user event packet;

When the user's dimension information changes, the user basic information carried in the user event packet includes the dimension of the user's current change. For the dimension that has not changed, the user basic information carried in the user event packet may not be included. If all the dimensions of the user are changed, the user basic information carried in the user event packet includes all the current dimensions of the user; if the basic information of the user carried in the user event packet includes the current change of the user Dimensions, only the changed dimensions of the saved basic information of the user are updated, and for the dimension that has not changed, no modification is made;

If the user's dimension does not change, the user carried in the user event packet is basically The information does not contain the dimension of the user. Correspondingly, after receiving the user event packet, the dimension information in the saved user basic information is not updated, and the dimension information of the current state of the user is directly obtained from the currently saved user information.

Step 102: Add the count value of the event-related indicator in the corresponding statistical object instance according to the dimension information and the event type of the current state of the user;

Specifically, the corresponding statistical object instance is searched according to the dimension information of the currently saved user state, and if the corresponding statistical object instance is found, the count value of the event related indicator in the statistical object instance is incremented by 1 according to the event type, if not Find the corresponding statistical object instance, create a new statistical object instance, and increase the count value of the event-related indicator in the statistical object instance by 1 according to the event type;

And searching for the corresponding statistical object instance according to the dimension information of the currently saved user state, specifically:

Calculating an index corresponding to the dimension information according to the dimension information of the currently saved user state; and searching for the statistical object instance corresponding to the index according to the index corresponding to the dimension information;

The dimension information of the user state includes specific information of multiple dimensions. In order to quickly find the instance of the statistical object corresponding to the dimension information, a hash algorithm or a message digest algorithm fifth edition (MD5) may be used. Or other algorithms for digital signatures first calculate an index corresponding to the dimension information, and then use an index to find a corresponding statistical object instance; the index may be a feature value;

A dimension information corresponds to a statistical object instance, where the statistical object instance includes specific information of the dimension and a count value of the event-related indicator; wherein, the counter can be counted by the counter; for the count value of the event-related indicator, The event type corresponds to the count value of a set of related indicators. A statistical object instance may include the count value of related indicators of various event types. In the actual application process, the event type that needs to be counted may be selected according to requirements, and then allocated according to the need. Counter of related indicators; For step 102, for example, the types of events to be counted are activation and paging. Correspondingly, there are two counters for activating related indicators in the statistical object instance, namely: a counter for counting the number of successful activations and a number of statistical activation failures. The counter, the counter of the paging related indicator also has two, namely: a counter for counting the number of successful paging times and a counter for counting the number of paging failures. When the event type in the received user event message is activated, the execution result is activated. On success, the value of the counter for the number of successful statistics activations in the corresponding statistical object instance is incremented by one.

And setting a dimension information according to the attribute of the user service, where the attribute of the user service includes a location area, a routing area, a service area, an access mode, an access point, and the like of the user, and one dimension information corresponds to one dimension space, and step 101 is performed. And 102, can aggregate multiple events occurring in the same dimensional space into one statistical record. For example, suppose the total number of events occurring in a statistical period is X, and the average number of occurrences of events in each dimensional space For P (P≥1), the number of statistical object instances is X/P. Therefore, the higher the number of events occurring in each dimension space, the fewer the number of statistical object instances. For example, for a network element, assuming that the number of events per second is 10000 in a statistical period, such as 15 minutes, the total number of events in 15 minutes is 10000*60*15=9000000, if each dimension space The average number of occurrences of the event is 100, and the corresponding statistical object instance will be: 9000000/100=90000, and the compression ratio is 1/100.

Of course, in the limit case, for example, the statistical period is very short. When the average number of occurrences of events in each dimension space is 1, that is, all events that occur are not repeated, the number of statistical object instances is equal to the number of events. For a network element, the number of events that occur in a typical unit time is fixed. Therefore, you can set the statistics period to control the total number of events to a predictable range.

The process of steps 101-102 may be referred to as a process of original statistics. After the original statistics are completed, according to the user's requirements, the statistical results are further analyzed and operated according to the user's attention object, and the operation result is output; wherein, the output mode may be Graphically, the user can visually see the result of the operation; the user needs to be an event statistics under a certain dimension of interest. As a result, it may also be an event statistical result or the like under the combination of dimensions.

During the actual application, you can set the statistics period. For example, set the statistics period to 15 minutes. After every 15 minutes, after the timer expires, save the statistics of the period to disk. Then clear all statistics instances and report the new statistics. The user event message restarts statistics, so that the user can further analyze and calculate the statistical result, so as to better meet the user's needs.

Embodiment 1

The application scenario of this embodiment is: statistics of the activation success rate of the SGSN.

The dimension information of the user in this embodiment includes: a location area identifier (LAI), a routing area identifier (RAI, a routing area identity), a service area identifier (S AI, a service area identity), an APN, and a wireless connection. Incoming technology (RAT, Radio Access Technology), and the name of the radio network controller (RNC, Radio Network Controller); events that need to be counted are attachment, activation, deactivation, and paging, etc., correspondingly configured for statistics a counter of the number of times of attachment success, the number of attachment failures, the number of activation successes, the number of activation failures, the number of successful deactivations, the number of failed failures, the number of successful paging attempts, the number of failed pagings, and the like, and a method for calculating an index corresponding to the dimensional information, For example: Hash algorithm, set the statistical period to 15 minutes.

After receiving the user event packet, the SGSN finds the saved user basic information according to the user identifier carried in the user event packet. If the user event packet carries the dimension of the event change, the SGSN carries the content according to the user event packet. The user information of the user information saved in the dimension update; if the dimension information is not carried, the dimension information in the saved basic information of the user is not updated; wherein the user event packet is an existing message, in this embodiment, the user event The message may be a message of an event such as attaching, activating, deactivating, and paging; when the user's geographic location, and/or access mode, and/or access point change, the user's dimensional information The change may occur. In this embodiment, specifically, when one or several of the LAI, LAI, RAI, SAI, APN, RAT, and RNC names are changed, the user's dimension information is changed. Chemical

Then, according to the dimension information of the currently saved user state, the index corresponding to the dimension information is calculated, and according to the index corresponding to the dimension information, the statistical object instance corresponding to the index is searched. If the corresponding statistical object instance is found, the statistical object is determined according to the event type. The count value of the event-related indicator in the instance is increased by 1. If the corresponding statistical object instance is not found, a new statistical object instance is created, and according to the event type, the count value of the event-related indicator in the statistical object instance is incremented by one. The statistics are saved every 15 minutes, and then all the statistics instances are cleared. The statistics of the newly reported user event packets are restarted. The original statistical result is shown in Figure 2. Only the activation is listed in Figure 2. The count value of the related indicator, namely: the number of activation successes and the number of activation failures.

Select the dimension or combination of dimensions of interest, such as: The selected dimension combination is (LAI=32432,

RAI=3256, SAI=3247, APN=xcom.net, RAT=UTRAN), the selected time period is: 16:00:00 to 18:15:00, according to the original statistical results, you can get as shown in Figure 3. The dimension combines the query results, and according to the query result, the activation success rate is obtained, and the output result as shown in FIG. 4 is obtained; in practical application, the dimension of interest can be selected by using Structured Query Language (SQL, Structured Query Language) or Dimensional combination, then calculate the activation success rate. Among them, the calculation formula of the statistical activation success rate is:

Activation success rate = number of activation successes / (number of successful activations + number of activation failures) *100; Here, the number of activation successes and the number of activation failures are the number of activation successes and activation failures of the dimension or dimension combination query results of interest at that time. frequency.

As can be seen from the above description, metric analysis and comparison of any dimension or combination of dimensions of the SGSN can be achieved.

Embodiment 2

The application scenario of this embodiment is: Statistics of the bearer update success rate of the Gateway General Packet Radio Service Support Node (GGSN).

The dimension information of the user in this embodiment includes: LAI, RAI, SAL APN, RAT, And the RNC name, etc.; the events to be counted are activated, deactivated, and bearer updated, etc., correspondingly, configured to count the number of successful activations, the number of activation failures, the number of successful deactivations, the number of deactivation failures, and the number of successful update attempts. A counter that carries the number of update failures, and sets a method for calculating an index corresponding to the dimension information, for example, a hash algorithm, and sets the statistics period to 15 minutes.

After receiving the user event packet, the GGSN finds the saved user basic information according to the user identifier carried in the user event packet. If the user event packet carries the dimension of the event change, the GGSN carries the content according to the user event packet. The user information of the user information saved in the dimension update; if the dimension is not carried, the dimension information in the saved basic information of the user is not updated; wherein the user event packet is an existing message, in this embodiment, the user event report The text may be a message that activates, deactivates, and carries updates, etc.; when the user's geographic location, and/or access method, and/or access point change, the user's dimensional information changes. In this embodiment, specifically, when one or several of the LAI, LAI, RAI, SAI, APN, RAT, and RNC names change, the user's dimension information is changed; The dimension information of the currently saved user state, the index corresponding to the dimension information is calculated, and then the index corresponding to the dimension information is used. Find the statistical object instance corresponding to the index. If the corresponding statistical object instance is found, add the count value of the event-related indicator in the statistical object instance to 1 according to the event type. If the corresponding statistical object instance is not found, create a new one. The statistical object instance, and according to the event type, the count value of the event-related indicator in the statistical object instance is incremented by 1, and the statistical result is saved every 15 minutes, and then all the statistical instances are cleared, and the newly reported user event message is re-renewed. Starting the statistics, the original statistical results obtained are similar to the original statistical results shown in Figure 2.

Select the dimension or combination of dimensions of interest, and according to the original statistical result, you can get the dimension or dimension combination query result, and then according to the query result, you can count the success rate of the bearer update. In actual application, you can use SQL to select the dimension or dimension of interest. Combine, then calculate Load update success rate. The calculation formula for the success rate of the bearer update activation is:

Bearer update success rate = number of successful update of the bearer / (the number of successes of the bearer update + the number of failed bearer updates) *100;

Here, the number of successes of the bearer update and the number of failures of the bearer update are respectively the number of successes of the bearer update and the number of failed bearer updates of the dimension or dimension combination query result in the period of time.

As can be seen from the above description, FFT analysis and comparison of any dimension or combination of dimensions of the GGSN can be achieved.

In order to implement the above method, the present invention further provides a user behavior statistics device based on a user event. As shown in FIG. 5, the device includes: a message analysis module 51 and a statistics module 52; wherein, the message analysis module 51, After receiving a user event message, the dimension information of the user's current state is obtained from the currently saved user basic information, and the obtained current dimension information of the user is sent to the statistics module 52;

The statistic module 52 is configured to: after receiving the dimension information of the current state of the user sent by the packet analysis module 51, accumulate the count value of the event-related indicator in the corresponding statistical object instance according to the dimension information and the event type of the current state of the user.

The device may further include a storage module 53 configured to store user basic information and a statistical object instance, where the user basic information includes dimension information of a current state of the user.

The device may further include a collection module 54 configured to receive a user event message, and when the user event message carries the dimension in which the event changes, the storage module 53 is updated according to the dimension carried in the user event message. Saved user basic information.

The statistic module 52 is specifically configured to: search for a corresponding statistical object instance according to the dimension information of the currently saved user state, and if the corresponding statistical object instance is found, according to the event type, the event related indicator in the statistical object instance is The count value is incremented by 1. If the corresponding statistical object instance is not found, a new statistical object instance is created, and the count value of the event-related indicator in the statistical object instance is incremented by one according to the event type. The device may further include: a timer and a clearing module; wherein

a timer, configured to trigger a clearing module after a timeout;

The clearing module is used to save the statistics after being triggered by the timer, and then clear all statistical object instances.

The storage module 53 is further configured to save a statistical result.

The device may further include: a calculation module, configured to further analyze and calculate the statistical result according to the user requirement after the statistics are completed, and output the operation result.

The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included. Within the scope of protection of the present invention.

Claims

Claim

A user behavior statistics method based on a user event, the method comprising: after receiving a user event message, obtaining dimension information of a current state of the user from the currently saved user basic information;

2. The method according to claim 1, wherein the method further comprises: before acquiring the dimension information of the current state of the user from the currently saved user basic information, the method further comprising:

The method according to claim 1 or 2, wherein the sum of the event-related indicators in the corresponding statistical object instance is accumulated according to the dimension information and the event type of the current state of the user, as follows:

According to the dimension information of the currently saved user state, the corresponding statistical object instance is searched. If the corresponding statistical object instance is found, the count value of the event related indicator in the statistical object instance is incremented by 1 according to the event type, if no corresponding corresponding is found. For the statistical object instance, a new statistical object instance is created, and the count value of the event-related indicator in the statistical object instance is incremented by one according to the event type.

The method according to claim 3, wherein the searching for the corresponding statistical object instance according to the dimension information of the currently saved user state is:

The method according to claim 1, wherein the method further comprises: after the timer expires, saving the statistical result, and clearing all the statistical object instances, and restarting the statistics.

6. The method of claim 1, 2 or 5, wherein the method further comprises:

A user behavior statistics device based on user events, characterized in that the device comprises: a message analysis module and a statistics module;

The device according to claim 7, wherein the device further comprises a storage module, configured to store user basic information and a statistical object instance, wherein the user basic information includes dimension information of a current state of the user.

The device according to claim 8, wherein the device further comprises a collection module, configured to receive a user event message, and when the user event message carries a dimension in which the event changes, according to the The dimension information carried in the user event packet updates the basic user information saved by the storage module.

The device according to claim 7, 8 or 9, wherein the device further comprises: a timer and a clearing module;

a timer, configured to trigger a clearing module after a timeout;

The clearing module is used to save the statistics after being triggered by the timer, and clear all statistical object instances.

11. Apparatus according to claim 7, 8 or 9 wherein the apparatus is further The calculation module is configured to further analyze and calculate the statistical result according to the user requirement after the statistics are completed, and output the operation result.