KR20090050405A - Method and apparatus for classifying user behaviors based on the event log generated from the context aware system environment - Google Patents

Abstract

The present invention relates to a method and apparatus for classifying a user's behavior using an event log that records events occurring in a context aware system environment. As information to be used to provide an active and intelligent service to a user in a context aware system environment, action items including a series of event actions as conditions are classified on the action classification table. The action classifier of the present invention determines which action item of the action item on the action classification table is associated with an event action occurring at a specific time, and then the subsequent event actions including the event action are defined as the condition of the action item. Calculate the probability of satisfying the event action). The probability thus calculated is appended to the associated event action and created as a behavioral probability list. The behavior probability list is learned by the learning engine, and the user behavior pattern generated by the learning is used to provide active and intelligent services to the user.

Description

METHOD AND APPARATUS FOR CLASSIFYING USER BEHAVIORS BASED ON THE EVENT LOG GENERATED FROM THE CONTEXT AWARE SYSTEM ENVIRONMENT}

The present invention relates to the classification of user actions based on event logs, and more particularly, to a method and apparatus for classifying user actions based on event logs occurring in a context aware system environment.

The present invention is derived from the research conducted as part of the IT growth engine core technology development project of the Ministry of Information and Communication [Task Management Number: 2006-S-026-02, Title: Development of URC server framework for active services].

Event logging is a standard way of recording software and hardware events on computer systems. The hardware module or software module of the computer system generates an event and sends it to an event logger, which stores the event in memory. In other words, an event log is a record of a series of actions of processes running on a computer system.

Event logs are mainly used to determine the cause of failures. For example, by analyzing the event log, you can detect conflicts between processes, hacks or infiltrations of viruses.

Recently, event logs have also been used to generate statistics such as the number of requests from a user for a particular application or the number of times a user connects to the system. This statistics generation function using the event log is particularly useful in a system providing a commercial service for an unspecified number.

Large Internet shopping sites, such as E-bay and Amazon, use event logs to analyze purchase correlations between books by user group, and then when a user wants to buy a particular book, He uses a personalization service technique that recommends other books.

As one of such online personalization service techniques, the user's propensity is analyzed based on event information generated by a plurality of users accessing an Internet website, and a plurality of category data suitable for the propensity is transmitted to the user's computer. The technique of outputting is known. This technology is suitable for services such as personalized information online and targeted marketing in sites that require personalization services such as large e-commerce sites or large information portal sites.

Meanwhile, a context aware system (Context Aware System) is a system for realizing ubiquitous computing in a limited space such as home, office, etc., and should be able to control all devices in a given space, and user in a given space. And the location of the user.

In a context aware system, a user has a user terminal that stores his own information and can communicate with the context aware system. The user terminal also has a display which displays the contents of communication with the situation recognition system. The context aware system recognizes the user and the location of the user by communication with the user terminal and displays a service suitable for the recognized information to the user. The user selects the desired service according to the displayed information.

For example, if a space is divided into several rooms, and a printer is installed in some of the several rooms, and the user wants to use the printer, the situation recognition system is configured to communicate with the user terminal. It recognizes the user and its location and informs the user of the location of the printer closest to the user through the display of the user terminal.

In addition, depending on when and where the user is, a situational awareness system is known in which inputs to the same button of the user terminal provide different services to the user.

Conventional event log utilization technology is used not only for analyzing the cause of failure using the event log, but also for a technique of increasing service satisfaction for an unspecified number of visitors visiting a specific web site using purchase tendency information based on a purchase history log. However, there is a limit in that it cannot reflect other historical information that is indirectly related to the purchase, such as purchase information obtained from surrounding people, or purchase information obtained from another medium such as a TV. In other words, on an open broadband network such as Internet, there is a limit in that it is impossible to consider all possible service environments surrounding a specific individual in order to improve the service satisfaction of the specific individual.

Since the context aware system operates within a limited space, it may provide a clue to solving the above problems of the conventional event log utilization technique, but the conventional context aware system may not provide a solution to the physical environment surrounding the user. Because it operates only under typical and preset conditions, there is a limitation that it cannot provide intelligent and active services that reflect the user's execution pattern for the service.

Accordingly, an object of the present invention is to learn a user behavior pattern reflecting a user's changing tendency based on an event log in a context aware system environment, and provide intelligent and active services to a user based on the learned user behavior pattern. To make it available.

Another object of the present invention is to learn a user behavior pattern by utilizing an event log of a service requested by a user in a context aware system environment, and based on the learned user behavior pattern, provide a service that is most suitable for a user's situation. To make it available.

In order to achieve the above object, the user behavior classification method of the present invention is a user behavior classification method based on an event log of an event occurring in a situation recognition system, and (a) a plurality of events each including a series of event operations as conditions. Classifying the user's action items into an action classification table; and (b) an event log of an event occurring in the context awareness system, the event log including user identity information, time information, and user location information related to the event. (C) generating an action vector indexing the created event log by user, by time, and by user location, wherein the action vector includes an event action by time; (d) the time-based event action of the action vector is associated with an action item in the action classification table. Determining whether there will be, (e) determining a probability that a series of event actions constituting a condition included in the determined action item will occur after the time-based event action of the action vector, and (f) the And generating an action probability list appending a probability to the time-based event action of the action vector.

In addition, in order to achieve the above object, the user behavior classification method of the present invention includes (a) generating an action vector indexing an event log of an event occurring in a situation recognition system by user, by time, and by user location. And (b) which action item in the action classification table includes a plurality of action items each including a series of event actions as conditions for the time-based event actions included in the action vector. (C) calculating a probability that a series of event actions constituting a condition included in the determined action item occur after the time-based event action of the action vector; and (d) the probability Generating a behavior probability list attached to the time-based event action of the behavior vector.

In this case, the behavior classification table may be prepared with reference to the behavior vector. And the behavior items and conditions in this behavior classification table may be determined by one or more of regular / repetitive behavior, irregular / intermittent behavior, behavior according to habit, and behavior according to propensity / symbol.

In addition, in order to achieve the object as described above, the user behavior classification apparatus of the present invention is a user behavior classification apparatus based on an event log of an event occurring in the situation recognition system, and is an event log of an event occurring in the situation recognition system. An event logger for generating an event log including user identity information, time information, and user location information related to the event, and an action vector indexing the created event log by user, time, and user location An action vector extractor, wherein the action vector includes a time-based event action, and a time-based event action of the action vector includes a plurality of action items each including a series of event actions as a condition. To any action item in the categorized action taxonomy Calculate a probability that a series of event actions constituting a condition included in the determined action item occur after the time-dependent event action of the action vector, and calculate the calculated probability as the It includes a behavior classifier for creating a behavior probability list attached to the time-based event action.

According to the present invention, a basis for extracting a behavior pattern of a user in a context aware system environment is provided based on a user behavior vector utilizing an event log.

Further, according to the present invention, an intelligent and active service according to the environment surrounding the user (for example, the user and the location of the user) based on the user's behavior pattern in the context aware system environment by utilizing the event log. Provides a foundation for automatically providing the user with

In addition, according to the present invention, based on the behavior pattern of the user extracted by the analysis of the event log, it provides a basis for intelligent and active provision of the service most suitable for the environment in which the user.

Prior to describing an embodiment of the present invention, the term 'event operation' herein includes time information, user identity information, user location information, device operation information, sensor information, and the like detected in a context aware system environment. Therefore, it should be understood as a concept that includes all operations of objects that can be detected in a context aware system environment. Accordingly, the operation of the device related to the time information of FIG. 6 and FIG. 8 and the corresponding time information is an 'event operation'. In addition, elements constituting 'conditions' of FIG. 7 may also be defined as 'event actions'.

Hereinafter, with reference to the accompanying drawings will be described an embodiment of the present invention.

The user behavior classification apparatus 10 according to the embodiment of the present invention shown in FIG. 1 includes an event log generator 11, an action vector extractor 12, an action classifier 13, and a storage 14. The various systems and functional modules described herein, including the user behavior classification device 10, may be incorporated into general hardware configurations, such as processor, memory, and I / O devices of a computer system and software configurations for performing the functions. Is implemented. In addition, in the present embodiment, the user behavior classification device 10 is depicted as a separate device from the situation recognition system 20 for the purpose of understanding and convenience of explanation. It should be understood that they may be integrated.

The user behavior classification apparatus 10 includes a situation recognition system 20 including four subsystems 20a, 20b, 20c, and 20d illustrated in FIG. 2, a learning engine 30, and a service recommendation engine 40. And a user terminal 50. The situation recognition system 20 controls the operation of a plurality of devices in accordance with an input from the user terminal 50, and the situation recognition system 20 further includes information (for example, user ID) grasped from the user terminal 50. And / or location information of the user), a list of suitable services is presented to the user, or a suitable service is automatically provided to the user.

2 shows four subsystems constituting the context awareness system 20 according to the present embodiment. The four subsystems include a home network subsystem 20a that controls and manages a plurality of devices, a user identification subsystem 20b that identifies who the user is, and a user location recognition subsystem 20c that recognizes the location of the user. ), And a sensor network subsystem 20d that controls and manages a plurality of sensors.

The home network subsystem 20a is a known system including a plurality of home devices including electronic appliances in a home, and a home server for integrally controlling and managing the plurality of devices by wire / wireless communication. The home network system controls the operation of the home device according to the device control signal from the user terminal. In addition, the home network system may be connected to the Internet or a mobile communication network. In this case, the user of the home network system may monitor a situation in the home or directly control a device in the home from the outside.

The user identification subsystem 20b includes a user terminal possessed by a user and stored with the user's unique information, and a server for verifying the user's identity based on the user's unique information received from the user terminal. to be.

The user location recognition subsystem 20c is a system capable of recognizing the position of an object within a limited area, such as a context aware system environment, such as a known RF-based location recognition system, an infrared-based location recognition system, and an ultrasound-based location recognition system. If anything, it may be.

The sensor network subsystem 20d is an underlying system for realizing ubiquitous computing, which is generated in a context aware system environment from physical sensors such as door sensors, window opening / closing sensors, gas sensors, fire sensors, pressure sensors, temperature sensors, and the like. It is a known system for detecting a physical situation.

On the other hand, the learning engine 30, the user behavior probability list (Fig. 7) stored in the storage 14 by the user behavior classification method according to an embodiment of the present invention, such as machine learning theory and neural network learning theory, Learn by computer learning theory to build user behavior patterns. The service recommendation engine 50 selects a user behavior pattern suitable for the event log received from the event builder 11 from the user behavior pattern established by the learning engine 40, and then, according to the selected user behavior pattern. The service most likely to be used by the user is presented to the user through the user terminal 50. The learning engine 30 and the service recommendation engine 40 enable intelligent and active service provision for users in the context aware system environment.

Since the operation of the learning engine 30 and the service recommendation engine 40 is known to be implemented by applying a known computer learning theory and the aforementioned conventional service recommendation technique, the learning engine 30 and the service recommendation engine 40 ) Will be omitted.

Hereinafter, the configuration and operation of the user behavior classification apparatus 10 according to an embodiment of the present invention will be described in detail.

The event logger 11 may include an event merger 11a, an event logger 11b, and an event filter 11c. The event merger 11a collects / integrates the events generated in each of the subsystems 20a, 20b, 20c, and 20d of the situation awareness system 20 to generate a raw event log for each subsystem, and generates an event logger. To 11b.

In Figure 3, examples of raw event logs are shown. The raw event log (H_LOG) for events that occurred in the home network subsystem includes the user ID (UserID), event name (EventName), event description (EventDescription), task ID (TaskID), task description (TaskDescription), and time stamp ( TimeStamp) may be recorded. The user ID is an identifier given to the user recognized by the user confirmation subsystem 20b. The event name and event description are names and descriptions uniquely given to the event. The task ID and the task description are identifiers and descriptions given to specific application applications or services executed by the application application of the context aware system.

A user ID (UserID), a location ID (LocationID), a zone ID (ZoneID), a time stamp (TimeStamp), and the like may be recorded in the raw event log (L_LOG) for an event occurring in the user location recognition subsystem 20c. . The location ID is an identifier given to the user's location recognized by the user location recognition subsystem 20c, and the area ID is an identifier given to the user's area recognized by the user location recognition subsystem 20c. 'Location' is wider than 'Area'. The time stamp is time information at which the position of the user is recognized.

The sensor ID (SensorID), the event name (EventName), the event description (EventDescription), the time stamp (TimeStamp), and the like may be recorded in the raw event log S_LOG for the event occurring in the sensor network subsystem 20d. The sensor ID is an identifier of a sensor recognized by the sensor network system 20d.

The event logger 11b stores the raw event log received from the event merger 11a in an internal temporary storage and delivers it to the event filter 11c. The event filter 11c is a final event in which the raw event log removes meaningless events (such as normal start / stop events of the system) or invalid events (such as channel change events when the TV is not turned on). Create a log and save it back to the temporary store. The event filter 11c is useful for reducing the processing load of the system.

4 shows examples of the final event log of this embodiment. The task execution log consists of a user ID (UserId), a task ID (TaskId), a task description (TaskDescription), an execution start time (InvokedTime), and a duration (Duration). It indicates how long it has been running continuously. The command execution / location log is composed of user ID (UserId), event name (EventName), event description (EventDesc), location ID (LocationId), zone ID (ZoneId), task ID (TaskId), and execution start time (InvokedTime). Configured, and for a particular user, indicates when and at what location (region) for a particular task for a particular event. The continuous execution task log consists of a user ID (UserId), task ID (TaskId), next task ID (NextTaskId), and execution start time (InvokedTime). For a specific user, continuous execution of a specific task and the next task is performed. Indicates when it started. In addition, the final event log may include events that occur in the sensor network subsystem 20d to determine user behavior patterns for the physical environment (light, sound, temperature, movement, etc.) Any event deemed necessary to extract the behavioral pattern of can be included in the final event log.

The behavior vector extractor 12 is driven at a predetermined time interval and consolidates / cleans up the final event logs stored in the temporary storage to generate the behavior vector in the service log format as shown in FIG. ). In addition, the behavior vector extractor 12 indexes the behavior vectors collected at a given time interval (for example, 'day') for each user, hourly, and place and stores them in the storage 14. 6 shows an example of an action vector indexed by user / time / place.

The behavior vector of FIG. 5 is referred to the preparation of the behavior classification table (FIG. 7) described later, and the indexed behavior vector of FIG. 6 is used for the creation of the behavior probability list (FIG. 8) described later. In addition, although the behavior vector of FIG. 5 is composed of eight vector elements, the behavior vector of FIG. 5 may further include other vector elements (eg, information detected by physical sensors) that are determined to determine a user's behavior pattern. have. In this case, the behavior classifier described later enables more diverse and sophisticated user behavior classification.

The behavior classifier 13 creates a behavior classification table as shown in FIG. 6 with reference to the behavior vector (FIG. 5) integrated / arranged by the behavior vector extractor 12, and stores it in the storage 14. The behavior categorization table shown in FIG. 7 shows the user's actions or peripheral device actions (ie, 'event actions') included in the series of events when the sequence of events occurs in the context aware system environment. It is a summary of what kind of action to classify.

In addition, the behavior classifier 13 creates and stores the behavior probability list as shown in FIG. 8 with reference to the behavior classification table of FIG. 7 and stores it in the storage 14. To this end, the action classifier 13 first determines which action item on the action classification table (FIG. 7) relates to the visual event action of the indexed action vector of FIG. For example, for an event action of "06:10 light2 on", an action item having a condition satisfying the time information of the event action is first selected, and then "light2 on" is selected from the selected action items. An action item having a condition satisfying the event action information is further selected.

If GetUp is 6:10 am in the condition of the action item of 'morning weather', in the above example, if the condition of "Time = GetUp" is exactly satisfied and the light (bedroom) corresponds to "light2", the " Since the condition of Light (bedroom) = on ”is met exactly, it may be determined that the event action of“ 06:10 light2 on ”of the above example is associated with an action item of 'morning weather' (A000). On the other hand, in the above example, the action item 'morning weather' is uniquely determined (the visual information alone does not provide a choice for other action items), but when a plurality of suitable action items are detected, all of the plurality of action items are selected. In addition, one action item may be selected according to a predetermined rule or the like.

The behavior classifier 13 then creates a behavior probability list that attaches the probabilities for the selected behavior item (s) to each event action. That is, the behavior probability list shown in FIG. 8 is a summary of the behaviors of the indexed behavior vectors of FIG. The behavior probability list is a stochastic summary of which behavior items are classified on the behavior classification table (FIG. 7) when a specific user performs a specific behavior at a specific time (large time). For example, in FIG. 8, the indication “06:10 light2 on (A0000: 0.5: NA: 0.5)” indicates that the act of turning on the light 2 at 6:10 am is the act of 'morning weather' on the act classification table (FIG. 7). It indicates that the probability of being linked to the content of a series of conditions of the item is 0.5 and the probability NA is 0.5. In addition, "06:15 window2 open" means opening window 2 at 6:15 am, which is not attached to a probability indication. This is because this action is an action or condition not previously defined on the action classification table (FIG. 7). Therefore, if the act of opening the window 2 is added to the act classification table (Fig. 7), the probability will be attached to the act.

This probabilistic theorem can be performed by, for example, the following probability distribution analysis method. For example, the behavior classifier may, when user A turns on the bedroom light between 6 am and 7:00 am each day, a subsequent series of user behaviors or peripheral actions, including that behavior, are displayed on the behavior classification table (FIG. 7). Calculate the probability of constructing a 'morning' behavior. That is, a total of 10 behavior probability lists have been accumulated up to 10 days (that is, a total of 10 days if the given time interval is 'day'), and among the 10 behavior probability lists, User A is 6 AM and 7 AM every day. When the bedroom light was turned on in the meantime, if there were seven behavior probability lists that satisfies all the conditions for subsequent user behaviors or peripheral behaviors to be classified as 'morning' behaviors in the behavior classification table (FIG. 7), If A acts to turn on the bedroom light between 6 am and 7:00 am today, the behavior classifier will satisfy the conditions attached to the 'morning phase' behavioral items on the behavior classification table (Figure 7) by that behavior. You can decide that the probability is 70%.

Of course, the actual probability calculation should be performed more complicatedly and precisely than in the above example, but the basic concept of calculating the posterior porbability distribution of the current behavior from previous accumulated knowledge is the same as the above example. have.

Hereinafter, the behavior classification table illustrated in FIG. 7 will be described in more detail.

Each action item in the action classification table consists of a code assigned to each action, a description of the action, and the conditions to be classified as the action. Of course, the action items can be added or deleted as necessary. In addition, the condition to be classified as the action is a condition for actions useful for learning of the learning engine 15 to be described later, and includes time, device, sensor, location, image recognition, and the like.

That is, the behavior classification table of FIG. 7 is a list of actions that are considered to be effectively reflected in service provision or service recommendation associated with learning. In addition, the behavior items and conditions in the behavior classification table of FIG. 7 are classified by referring to regular / repetitive behavior, irregular / intermittent behavior, behavior according to habit, behavior according to disposition / symbol, and expected variables for each behavior. This can be obtained based on the following basic information.

-Behavior pattern = given environment information + user profile + service information

-Given environmental information = place + time + schedule + past behavior history

-User profile = demographic information + preference

-Anthropological information = gender + age + occupation + religion + other information

-Preference = personality type (e.g. MBTI) + other information

Service information = list of service information that can be executed in a given situation

The given environmental information can be grasped as follows, for example, from the behavior vector of FIG. The location information can be grasped by the location ID (LocationID) or the like of the behavior vector of FIG. 5. Time information can be grasped | ascertained from execution start time (InvokedTime), duration (Duration), etc. of the action vector of FIG. The schedule information can be grasped from the accumulated time information. The past behavior history can be grasped from accumulated place information, time information, and schedule information.

Information such as the user profile, anthropological information, and preference may be grasped by, for example, a user ID of FIG. 5. The user ID is associated with various personal information about the user, and since the personal information is previously input to the system by the user, the user profile can be identified by the user ID.

The service information can be grasped by the task list (TaskID) of the action vector of FIG. 5 and the service list predefined by the system (when it is not enough to grasp the service information by the accumulated task ID only).

On the other hand, the following is an example of the regular / repetitive behavior that is used to determine the action items and conditions to be included in the behavior classification table of FIG. In the following examples, referring to the variable related to 'morning weather', a series of actions of a particular user sleeping in bed, waking up at a specific time, turning on the bedroom light, and exiting the living room is described as' It can be defined as an action item of 'morning weather'. If such a typical set of behaviors are classified into specific behavioral categories (FIG. 7) and made into a behavioral probability list (FIG. 8), a good learning effect by the learning engine 30 can be expected (better from typical preceding behaviors). Learning outcomes can be expected, and if typical behavioral patterns are unpredictable or inadequate, it is unclear which behaviors to learn, so good learning outcomes cannot be expected). The behavioral classification for the other actions below is done in the same way.

-Morning weather

  -If you have a bed, go to bed.

  -Office workers usually have a constant waking time.

  -After waking up, usually turn on the bedroom lights and go out to the living room.

-Evening bedtime

  -Bedtime is constant (or irregular).

  -When you go to bed, you usually sleep in bed.

-Going to work (going out)

  -Time for work is fixed and regular (or irregular).

  -If the weather is bad (rain, snow, etc.), you will be late for work.

  -If you are in poor condition, you will be late for work.

  -Make sure you go out on your anniversary.

  -Do not go out if the weather is bad.

Off Work

  -The time for leaving work is fixed and regular (or irregular).

  -If the weather is bad (rain, snow, etc.), leaving home will be delayed.

  -Return home relatively late on the anniversary.

- rest

- shopping

-Cleaning / washing

  -(Most workers) clean up (while listening to music) on holidays.

- light on

  -Turn on the lights when it is dark when you are indoors.

In addition, there may be examples of irregular / intermittent behavior referred to for determining behavior items and conditions to be included in the behavior classification table of FIG. 7.

-During the holiday season, search a lot of tourist information (increasing interest).

-I watch sports news mainly during the World Cup season.

-I see a lot of political news in election season.

-People who catch a cold often turn on the humidifier indoors.

In addition, there may be examples of behavior according to a habit, which are referred to for determining behavior items and conditions to be included in the behavior classification table of FIG. 7.

-Sleep late on weekends (Saturday, Sunday, national holidays, etc.).

-I read a book and sleep before going to bed.

-Sitting on the couch in the living room, always watching TV or listening to music.

-If you wake up in the morning, listen to the news.

-(Most workers) clean up (while listening to music) on holidays.

-If the room is hot, turn on the air conditioner.

-If the room is not hot but the humidity is high, turn on the dehumidifier.

-If the room is cold, turn on the heater.

-(Workers) check e-mails on holidays (weekends, etc.).

In addition, there may be examples of the behavior according to the propensity / symbol, which is referred to to determine the action items and conditions to be included in the action classification table of FIG. 7.

-The younger you are, the more sensitive you are to fashion.

-I read a book and sleep before going to bed.

-If there are other people around you, don't watch TV or watch programs that people around you want.

-Bedtime varies from person to person.

-I like to listen to the same piece of music repeatedly (to some extent).

-On cloudy days (rain, frowning), I listen to sad music / dance music mainly.

The behavior classification table (FIG. 7) and the behavior probability list (FIG. 8) stored in the storage 14 are used by the learning engine 30 and the service recommendation engine 40 as described above, and the situation recognition system environment. It allows you to provide active and intelligent services to your users.

9 is a flowchart of a user behavior classification method according to an embodiment of the present invention.

The event merger 11a collects / integrates the events generated in each of the subsystems 20a, 20b, 20c, and 20d (S501), and creates a raw event log (S503). The event filter 11c removes a meaningless event or an invalid event from the original event log (S505), and creates a final event log (S507). The behavior vector extractor 12 consolidates / organizes the final event logs stored in the temporary repository, generates a behavior vector in the service log format as shown in FIG. 5 and stores it in the repository 14, Behavior vectors collected at time intervals (eg, 'day') are indexed by user / time / place (FIG. 6) and stored in the storage 14 (S509). The behavior classifier 13 refers to the behavior probability list (FIG. 7) with reference to the behavior classification table (FIG. 7) created in the storage 14 with reference to the behavior vector (FIG. 5) integrated / arranged by the behavior vector extractor 12. 8) is created and stored in the storage 14 (S511).

On the other hand, the present invention may be implemented by a computer-readable recording medium recording a program for executing the user behavior classification method.

Although the embodiments of the present invention described above have been specified by specific configurations and drawings, it is intended to be clear that such specific embodiments do not limit the scope of the present invention. Accordingly, it is to be understood that the invention includes various modifications and equivalents thereof without departing from the spirit of the invention.

1 is a block diagram of a user behavior classification apparatus according to an embodiment of the present invention.

2 is a block diagram of a situation recognition system according to an embodiment of the present invention.

3 is an exemplary diagram of a raw event log in accordance with an embodiment of the present invention.

4 is an exemplary diagram of a final event log in accordance with an embodiment of the present invention.

5 is an exemplary diagram of an action vector in service log format according to an embodiment of the present invention.

FIG. 6 is an exemplary diagram of indexing by user / time / place for the behavior vector of FIG. 5; FIG.

7 is an exemplary diagram of an activity classification table according to an embodiment of the present invention.

8 is an exemplary diagram of a behavior probability list according to an embodiment of the present invention.

9 is a flowchart of a user behavior classification method according to an embodiment of the present invention.

Claims (5)

A method of classifying user behavior based on an event log of events occurring in a context aware system. (a) classifying a plurality of user action items each including a series of event actions as conditions into an action classification table; (b) creating an event log for an event occurring in the context aware system, the event log comprising user identity information, time information, and user location information related to the event; (c) generating an action vector indexing the created event log by user, time, and user location, wherein the action vector includes an event action by time; (d) determining which action item in the action classification table the time-based event action of the action vector is associated with; (e) determining a probability that a series of event actions constituting a condition included in the determined action item will occur after the time-based event action of the action vector; (f) generating a behavior probability list appending the probability to the time-based event action of the behavior vector. (a) generating an action vector indexing an event log for each event occurring in the context aware system by user, time, and user location; (b) determining which time-based event actions included in the action vector are associated with which action items on the action classification table in which a plurality of action items each containing a series of event actions as conditions is classified; , (c) calculating a probability that a series of event actions constituting a condition included in the determined action item will occur after the time-based event action of the action vector; and (d) generating a behavior probability list appending the probability to the visual event action of the behavior vector. The method according to claim 1 or 2, wherein the behavior classification table is created with reference to the behavior vector. The method according to claim 3, wherein the behavior items and conditions on the behavior classification table are determined by one or more of regular / repetitive behavior, irregular / intermittent behavior, behavior according to habit, and behavior according to propensity / symbol. How to classify user behavior. An apparatus for classifying user behavior based on an event log of events occurring in a context aware system, An event log for an event occurring in a context aware system, comprising: an event logger for creating an event log including user identity information, time information, and user location information related to the event; An action vector extractor for generating an action vector indexing the created event log by user, time, and user location, wherein the action vector includes an event action by time; The time-based event action of the action vector determines which action item on the action classification table in which a plurality of action items each including a series of event actions as conditions is associated with and is included in the determined action item. Calculating a probability that a series of event operations constituting a condition will occur after the time-based event action of the action vector, and generating a behavior probability list appending the calculated probability to the time-based event action of the action vector. A user behavior classification device comprising a behavior classifier.

Images