KR20100038272A - Method, apparatus, system and computer-readable media for determining user interest - Google Patents

Abstract

PURPOSE: A method, an apparatus, a system and a computer-readable media for determining user interests are provided to adaptively determine the preference and interests depending on a situation on the basis of a context-dependent recommendation. CONSTITUTION: A data item storage unit stores a data item, and the data item includes not only the interest field that has rating for user's interests but also the context field that has the information related to the context which concerns the rating. A grouping unit groups the data items into environment groups. A partition generating unit generates a first partition having a context group having the data items over a first critical value. A clustering unit averages the ratings for the interests in the data items which belong to the context group within the first partition, and performs the clustering operation for the each context group in the first partition.

Description

Method, Apparatus, System and computer-readable media for determining user interest}

It relates to software application techniques related to situation-dependent recommendations based on clustering.

Conceptually, a computing system (e.g., computing device, personal computer, laptop, smartphone, mobile phone, etc.) receives a set of instructions (e.g., content or data) that describes how to process the information by receiving it and adjusting it. Or a computer program). Typically, information used by a computing system is stored in computer readable memory in digital or binary form. Complex computing systems are also capable of storing content containing their computer programs. The computer program may be built in a computer (or computing) device as a logic circuit fixed or provided on a microprocessor or computer chip. The computing system can be a support system that manages various resources (eg, memory, auxiliary devices, etc.) and services (eg, basic functions such as file open) and allows resources to share multiple programs. Such support systems are well known as operating systems (OSs) that provide programmers with an interface used for access to resources and services.

Today, various kinds of computing devices are used. These computing devices range in size, price, storage and processing performance, from expensive, high-performance servers, relatively inexpensive personal computers and laptops, to relatively inexpensive microprocessors or computer chips for storage, automobiles, and consumer electronics. Consideration exists over a wide range.

Recently, computing systems have become smaller and more portable. As a result, a variety of mobile and portable devices have become available. For example, wireless telephones, media players, personal digital assistants (PDAs), and the like, have emerged. In general, a portable device may be a pocket sized computing device. Such devices typically use small display screens and miniaturized keyboards. For example, in the case of a PDA, a touch screen interface capable of input and output simultaneously may be used.

In particular, portable communication devices (e.g., smart phones) have become so popular that almost everyone has one, and it is possible for some portable communication devices to provide a computing environment similar to that provided in a typical PC. Such smartphones are capable of providing application service developers with a complete operating system as a standardized interface and platform. Detailed communication regarding the popularization of portable communication devices will be described later.

In general, mobile phones or cell phones may be various portable electronic devices used in mobile communication. In addition to the voice transmission function used in general telephones, mobile phones have recently been able to provide various services. For example, such a mobile phone is capable of providing a short message service (SMS) for text messaging, e-mail, packet switching for Internet connection, a multimedia messaging service (MMS) for sending and receiving pictures and videos, and the like. Most mobile phones are capable of connecting to a network of base stations and can be connected to each other via a public switched telephone network (PSTN).

SMS is used as a means for sending and receiving short messages in such mobile phones. SMS was initially defined as part of the GSM series standard in 1985 as a messaging means for sending and receiving 160 text messages in a global system for mobile communication (GSM) mobile handset. Since then, service support has expanded to include alternative mobile standards such as ANSI CDMA networks and digital AMPS, satellite and terrestrial line networks. Most SMS messages are mobile-mobile text messages, and broadcasting messages can also be supported. The term SMS is often used as a non-technical term to refer to text messages in non-English-speaking Europe where GSM systems are well established.

MMS is a standard for more modern messaging systems that enables the sending and receiving of messages that include multimedia objects such as images, audio, and video in addition to text in SMS. MMS has been developed in cellular networks along with other messaging systems such as SMS, mobile instance messaging and mobile e-mail, and most of the standardization process has taken place in 3GPP, 3GPP2 and open mobile alliance (OMA).

Personalization of user experiences has been the focus of recent research. Personalization offers many benefits. For example, the advertisement provider may provide the advertisement based on personal preference or interest through personalization technology. In the case of an mp3 provider, it is also possible to provide a music list in consideration of individual music propensity. As such, through personalization, it is possible for some service providers to provide personalized services based on user preferences or interests.

But preference is not a static concept. In other words, user preferences are constantly changing. This change may be due to the current state or situation of the user. For example, preferences in the office and at home may be different. A common approach to preferences that allows users to explicitly specify preferences can be through a user profile or similar structure. However, some users do not like specifying preferences and have a problem that the preferences are not completely maintained.

In the present specification, a technique for enabling an adaptive determination of interest of a user's preference or interest in accordance with a changing environment is disclosed.

A method of determining a user's interest in accordance with an aspect of the present invention is related to a usage pattern of the user, the interest portion having at least one rating indicating the degree of interest the user has with respect to at least one or more interests and the rating Storing a data item comprising an environment portion having information about an environment associated with the environment; Grouping the data items into an environment group having a data item and an associated environment portion; For each environment group, determining whether the number of data items in the environment group is greater than or equal to a first threshold value; Creating a first partition having an environment group having a number of data items above the first threshold value; Averaging ratings for interests of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; And comparing the current environment with an environment group in the first partition to obtain user interest.

In addition, a method of determining a user's interest in accordance with another aspect of the present invention relates to a user's usage pattern and includes at least one interest portion having at least one rating indicating a degree of interest the user has for at least one or more interests; Storing a data item comprising an environment value comprising a time of day and including an environment portion having information about the environment associated with the rating; Grouping the data items into an environment group having a data item and an associated environment portion; For each environment group, determining whether the number of data items in the environment group is greater than or equal to a first threshold value; Creating a first partition having an environment group having a number of data items above the first threshold value; Averaging ratings for interests of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; Filtering, for the environment group of the first partition, an environment group from the first partition when the number of data items is greater than or equal to a third threshold and the number of statistically significant average ratings is less than or equal to a second threshold; And comparing the current environment with an environment group in the first partition to obtain an interest of the user.

An apparatus for determining interest of a user according to an aspect of the present invention includes an interface; And a plurality of processors, the processor being associated with a usage pattern of the user, the processor including an interest portion having a rating of the degree of interest the user has and an environment portion having information about the environment related to the rating. Store the data item; Group the data item into environment groups having a data item associated with the environment portion; For each environment group, determine whether the number of data items in the environment group is equal to or greater than a first threshold value; Create a first partition having an environment group having a number of data items above the first threshold; Averaging ratings for interests of data items belonging to environment groups in the first partition, clustering each environment group in the first partition; It is possible to obtain an interest of the user by comparing a current environment with an environment group in the first partition.

A system for determining a user's interest in accordance with an aspect of the present invention includes an interest portion and at least one rating associated with a usage pattern of the user and having at least one rating indicating a degree of interest the user has for at least one or more interests. Means for storing a data item including an environment portion having information about an environment associated with the environment; Means for grouping the data items into an environment group having a data item and an associated environment portion; Means for determining for each environment group whether the number of data items in the environment group is greater than or equal to a first threshold value; Means for creating a first partition having an environment group having a number of data items above the first threshold; Means for averaging ratings of interest of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; And means for obtaining a user interest by comparing a current environment with an environment group in the first partition.

According to the disclosed subject matter, statistically significant data items are clustered, so that it is possible to adaptively determine a user's interest in a changing environment.

Hereinafter, specific examples for carrying out the present invention will be described in detail with reference to the accompanying drawings.

According to one embodiment of the invention, the software application and other items are automatically recommended to the user based on data related to the current context and past usage information. It is possible. The context may refer to a situation or a background to which a user and / or a device used by the user belong. For example, the environment may indicate the location of the user, that is, whether the user is at home or in the office. The environment may also include the time of day. For example, the time of day may be morning, afternoon, evening, or the like. However, the environment is not limited to these, and in addition, it is possible to include various environment or situation information.

One way for the system to provide a preference is to have the user specify his or her preferences and use this preference to guide the application or service recommendation. For example, a user may specify a preference for software related to the business, and the system may use the preference to recommend software related to the business upon request. Thus, if there is a choice between a game application and a word processing application, the system is able to recommend a word processing application.

Alternatively, the user needs to train the device over a period of time. This can be seen as supervised training, which is not appropriate because it takes a lot of time and effort to train the device. Another drawback of the existing approach is that it does not adapt adaptively to changing preferences, circumstances, or habits of using the device. Thus, there is a need for improved technology for mobile devices.

According to an embodiment of the present invention, preference information for the user may be calculated through the current environment using the user's preference from another environment.

Data related to past usage information can be collected and stored at data points. The data point information can be stored as a vector. 1 illustrates a vector for storing environment and interest information about a single data point according to an embodiment of the present invention. In FIG. 1, context information includes a time 100, a location 102, and a temperature 104. Interest information represents the user's interest in a particular subject of interest and may include pop music 106 and classical music 108.

Each data point can be visualized as a point on a two-dimensional graph. 2 illustrates a graph according to an embodiment of the present invention. In the graph, the distance between data points represents the similarity level of the vector. For example, the data points can be grouped into clusters of similar data points. Clustering is the categorization of an object into different groups, dividing a data set into subsets (i.e. clusters) so that the data in each subset share a common property. It can be a making process. Such clusters can be used to allow the system to select the appropriate application for recommendation.

One method of clustering data points may be an attempt to determine the optimal number of clusters for given data points. This optimal number of clusters can be expressed as k. At this time, the method for determining the optimal number of clusters is beyond the scope of the present invention according to the present embodiment can be used a general method. However, the general method has several drawbacks. k must be determined dynamically, which requires significant processing power. Processing power may be lacking in mobile devices. In addition, it is difficult to determine the correct k, and inaccurate estimates have a negative impact on clustering and recommendation. Given a partitioned cluster, one method for pattern extraction is to compare the current data point with each cluster centroid and determine which cluster the current data point belongs to. It is then possible for the pattern of interest to be extracted from this cluster. For n-dimensional structures, the center can be the intersection of all hyperplanes that divide X into two parts of the same moment. This can be seen as the "average" of all points of X. Thus, the center of the cluster can be the average of all points in the cluster. 3 illustrates a graph of clustered data points according to an embodiment of the present invention. In FIG. 3, the center of 300 clusters may be 302 and the center of 304 clusters may be 306.

Apart from clustering, another way of extracting the pattern may be to group data points adjacent to the current data point. Adjacent data points may have an environment and interest that are close to those of the current data point. The pattern of interest can be extracted from these adjacent data points. However, this method must be compared between all data points, and thus may not be suitable for resource-constrained online calculations, such as mobile devices.

According to an embodiment of the present invention, the user's preference may be specified by a rating provided to various keywords. For example, if the user shows a strong interest in classical music and rock music, high ratings can be given to the keywords "classic music" and "rock music". As one example, the rating may have an integer value between 0 and 5. Rating 0 may indicate no interest and rating 5 may indicate strong interest. The user may assign ratings to only a few keywords and leave the remaining keywords unrated.

If a given environment and interest rating are stored in a log, this information can be vectorized. For the environment type, 1-in-N encoding can be used. This means that if the environment type is any one of N values, the environment type can be represented by a vector of size N. The position of the vector may correspond to the current environment being set to a value of 1 and the rest to a value of zero. 4 illustrates a context set according to an embodiment of the present invention. In FIG. 4, possible time environments (eg, breakfast, lunch, dinner) may be represented by corresponding values (eg, 1, 2, 3 respectively).

5 illustrates a context encoding according to an embodiment of the present invention. In FIG. 5, breakfast, lunch, and dinner can be encoded into a vector containing three items.

For each keyword in the instance vector, "x" may be used to specify an unrated keyword, and 0 through 5 may be used to specify a user interest level of the rated keyword. 6 illustrates an instance vector according to an embodiment of the present invention. In FIG. 6, it is possible that if the user is in the office in the morning, the classical music has a "5" rating and the pop music has no rating. And if the user is at home in the evening, it is possible for the user to give the pop music a "1" rating and not the classic music.

Given encoded data, the entire log is encoded using a combination of environment values. For example, all instances of "morning, office" can be grouped into one vector and all instances of "morning, home" can be grouped into another vector. During the grouping, the ratings for each keyword can be combined by calculating the average value of the ratings for the keywords. At this time, the rating given by the value "x" can be ignored. 7 illustrates grouped instance data according to an embodiment of the present invention.

The purpose of grouping is to find patterns based on statistical statistical significance. For example, if there are multiple data instances within an environment group, this may mean that a statistically significant environment pattern can be extracted from the environment group. The first threshold can be used to assign an environment group to two partitions. The first partition may include an environment group in which the number of data instances is equal to or greater than the threshold. The second partition may include an environment group in which the number of data instances is less than the threshold. For example, when the first threshold value is set to 5, the first partition may include an environment group having five or more data instances. In this case, the first threshold value may be dynamically used according to an administrator or an application to be executed.

As mentioned above, the rating data also includes a valid rating rating and a missing rating. Such missing ratings should be treated as unknown. If there are only a few valid ratings for a particular interest, the system will not be able to conclude any reliable rating for that interest. This is true even if the number of data instances in the group exceeds the first threshold. Therefore, it is necessary to filter out environmental groups. For each environment group in the first partition, the average rating value for the keyword is calculated through the effective rating for the keyword from all data instances of the environment group. The average number of ratings in the group is then calculated. If the average rating number is smaller than the second threshold value, the number of ratings in the environmental group may be considered to be statistically insignificant.

In addition, for each average rating, the system must consider how many effective ratings are averaged. If the average rating is calculated from an insufficient number of effective ratings, the average rating value is also not statistically significant. Therefore, define a third threshold value. The third threshold may be the minimum value of the number of effective ratings required for the average rating to be statistically significant.

With these three thresholds, instances are reduced to enough instances to extract meaningful patterns of interest. Such a meaningful partition may mean an environment group in which the number of data instances is greater than the first threshold, the average rating is greater than the second threshold, and the number of valid ratings is greater than the third threshold. have.

The environmental state and average rating in the selected environment group (ie, the first partition) are used as the initial center for the environmental state and rating. Thus, the initial k value is equal to the number of identifiable patterns of statistical significance (ie, the number of environmental groups of the first partition). The remaining data instances in other environment groups (ie, second partitions) are allocated based on this initial group. The remaining data instances can be represented as outliers.

The outlier may be assigned to an environment group of the first partition by calculating an Euclidean distance between the environment state and the environment state of each existing (in the first partition) environment group. One with the minimum Euclidean distance can be an environment group to which a new data instance is assigned. Once assigned, the assigned environment group can update its environmental state center by combining the newly allocated data instances.

Once an instance is assigned, the rating center of the environment group can be updated by recalculating the average rating for each keyword from all data instances in the group.

In one embodiment of the present invention, in addition to placing an outlier in an existing cluster, it is also possible to form a new "outlier cluster" that includes all outliers. In this way, computational power can be reduced.

When acquiring the user's preferences under the current environment, first obtain the current environment. The 1-N encoding is used to convert the current environment into the encoded environment. The current encoded environment is compared with the center of each cluster environment. It is possible that a cluster having the smallest distance between the cluster centers from the current environment is used, and the rating center of the selected cluster is used as the user's preference.

When a new data instance arrives (ie, a user indicates a new preference), it is possible for the system to place the new data instance in an existing cluster. If there are enough data instances in an existing cluster that can be grouped with new data instances, it is possible that the cluster is split into two. In one example, this is done through a single iteration. However, the evaluation of existing clusters may be performed repeatedly so that all assumptions made during the initial cluster calculation may be reevaluated. For example, the system may run the entire cluster algorithm again each time a new data instance is received.

8 illustrates a method of determining interest of a user according to an exemplary embodiment. At 800, a data item associated with a user's usage pattern is stored. At this time, the data item may include an interest portion having one or more ratings as the degree of interest the user has for one or more interests and an environment portion having information as an environment related to the rating. At 802, data items are grouped into environment groups. Each environment group is capable of having data items associated with the environment part. The environmental portion of a data item may be considered "related" if it has at least one environmental value in common. At 804, it is determined whether for each environment group the number of data items in the environment group is equal to or greater than the first threshold. At 806, a first partition is created. It is possible for the first partition to have an environment group with a number of data items equal to or greater than the first threshold. At 808, the rating for the interest in the data item in the environment group in the first partition is averaged, such that each environment group in the first partition forms a cluster. At 810, the environmental group is filtered from the first partition if the statistically significant average rating is less than the second threshold. In this case, the average rating may be statistically significant when the number of data items is equal to or greater than the third threshold value. At 812, data items of the environment group that do not belong to the first partition (i.e., data items not originally located at 806 or filtered at 810) are assigned to the cluster. Details thereof will be described later. At 814, the user's attention is obtained by comparing the current environment with an environment group in the first partition. This may include determining the maximum proximity cluster obtained by determining the Euclidean distance between the center of the cluster and the current environment. The average rating of the closest cluster can be used to recommend to users with a value.

9 illustrates a method of allocating data items of an environment group not included in the first partition according to an embodiment of the present invention. This may correspond to step 812 of FIG. 8. At 900, the Euclidean distance between the environmental portion of each data item of the environmental group not included in the first partition and the center of the cluster of the first partition is measured. At 902, for each data item of an environment group not included in the first partition, a data item is added to the cluster of the first partition having a minimum Euclidean distance between the center of the cluster and the environment portion of the data item. At 904, the center point for the cluster of the first partition is recalculated.

10 illustrates a method of allocating data items of an environment group not included in the first partition according to another embodiment of the present invention. This may correspond to step 812 of FIG. 8. At 1000, all data items of the environment group not included in the first partition are merged into one outlier cluster. At 1002, the centroid for the outlier cluster is calculated.

11 illustrates a system according to an embodiment of the present invention. In FIG. 11, the system may be, but is not necessarily limited to, a mobile device. For example, the system may include a mobile device having a processor that performs the foregoing process, and may also include a server that performs this process. The system may also include a desktop computer.

The system may include an interface 1100 and at least one processor 1102. The processor 1102 may perform the process described with reference to FIGS. 8 to 10. In addition, if the system includes a mobile device, the system may also include a load detection module 1104. The load detection module 1104 is capable of measuring the load level of the processor 1102 (ie, indicating how busy the processor is). In such a case, it is possible for multiple processors to perform storage, clustering, determination, etc. when the device is not used or relatively small, and to perform selection, calculation and recommendation when there is a real recommendation request by the user.

Meanwhile, embodiments of the present invention may be implemented in computer readable codes on a computer readable recording medium. Computer-readable recording media include all types of recording devices that store data that can be read by a computing system.

Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, which may also be implemented in the form of carrier waves (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computing systems so that the computer readable code is stored and executed in a distributed fashion. In addition, functional programs, codes, and code segments for implementing the present invention can be easily deduced by programmers skilled in the art to which the present invention belongs.

In the above, the specific example for the implementation of the present invention has been described. The above-described embodiments are intended to illustrate the present invention by way of example and the scope of the present invention will not be limited to the specific embodiments.

1 illustrates a vector for storing environment and interest information about one data point according to an embodiment of the present invention.

2 illustrates a graph according to an embodiment of the present invention.

3 shows a graph with clustered data points according to one embodiment of the invention.

4 illustrates an environment set according to an embodiment of the present invention.

5 illustrates environment encoding according to an embodiment of the present invention.

6 illustrates an instance vector according to an embodiment of the present invention.

7 illustrates grouped instance data according to an embodiment of the present invention.

8 illustrates a method of determining a user's interest according to an embodiment of the present invention.

9 illustrates a method of allocating data items of an environment group not included in the first partition according to an embodiment of the present invention.

10 illustrates a method of allocating data items of an environment group not included in the first partition according to another embodiment of the present invention.

11 illustrates a system according to an embodiment of the present invention.

Claims (20)

A data item associated with a usage pattern of the user, the data item including an interest portion having at least one rating indicating a degree of interest the user has for at least one interest and an environment portion having information about the environment associated with the rating; Storing; Grouping the data items into an environment group having a data item and an associated environment portion; For each environment group, determining whether the number of data items in the environment group is greater than or equal to a first threshold value; Creating a first partition having an environment group having a number of data items above the first threshold value; Averaging ratings for interests of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; And Comparing the current environment with an environment group in the first partition to obtain user interest. The method of claim 1, Filtering an environment group from the first partition when the number of statistically significant average ratings is less than or equal to a second threshold value for the environment group of the first partition, Wherein said average rating determines a user's statistically significant interest when the number of data items is greater than or equal to a third threshold. The method of claim 1, And encoding the data item into a vector. The method of claim 3, wherein Wherein each encoded vector comprises an environment portion and an interest portion, wherein the interest portion comprises a keyword list and one or more ratings corresponding to the keyword. The method of claim 1, Creating a second partition having an environment group having a number of data items that is greater than or equal to the first threshold value; And For each data item in an environment group in the second partition, the data item has the first partition having a minimum Euclidean distance between the center of the cluster in the first partition and the environment portion of the data item. Adding to the cluster within the method. The method of claim 1, Creating a second partition having an environment group having a number of data items that is greater than or equal to the first threshold value; And Merging all of the environment groups in the second partition into one outlier cluster. The method of claim 5, After adding the data item to a cluster in the first partition, recalculating centroids for each cluster in the first partition. The method of claim 1, Receiving a new data item; Placing the new data item in a cluster in the first partition; And If the cluster is about to have two clusters having a number of data items equal to or greater than the first threshold value, dividing the cluster in which the new data item is located. The method of claim 1, Wherein each environment portion has a plurality of environment values, each environment value representing a description of a characteristic associated with a state including a usage pattern. A preference value associated with a usage pattern of the user, the preference value having at least one rating indicating the degree of interest the user has for at least one interest and an environment value including time of day; Storing a data item comprising an environment portion with information; Grouping the data items into an environment group having a data item and an associated environment portion; For each environment group, determining whether the number of data items in the environment group is greater than or equal to a first threshold value; Creating a first partition having an environment group having a number of data items above the first threshold value; Averaging ratings for interests of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; Filtering, for the environment group of the first partition, an environment group from the first partition when the number of data items is greater than or equal to a third threshold and the number of statistically significant average ratings is less than or equal to a second threshold; And And comparing the current environment with an environment group in the first partition to obtain the user's interest. interface; And Includes multiple processors, The processor comprising: Store a data item associated with a usage pattern of the user, the data item including an interest portion having a rating of the degree of interest the user has and an environment portion having information about an environment related to the rating; Group the data item into environment groups having a data item associated with the environment portion; For each environment group, determine whether the number of data items in the environment group is equal to or greater than a first threshold value; Create a first partition having an environment group having a number of data items above the first threshold; Averaging ratings for interests of data items belonging to environment groups in the first partition, clustering each environment group in the first partition; And compare the current environment with an environment group in the first partition to obtain the user's interest. The method of claim 11, And the device is a mobile device. A data item associated with a usage pattern of the user, the data item including an interest portion having at least one rating indicating a degree of interest the user has for at least one interest and an environment portion having information about the environment associated with the rating; Means for storing; Means for grouping the data items into an environment group having a data item and an associated environment portion; Means for determining for each environment group whether the number of data items in the environment group is greater than or equal to a first threshold value; Means for creating a first partition having an environment group having a number of data items above the first threshold; Means for averaging ratings of interest of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; And Means for obtaining a user interest by comparing a current environment with an environment group within the first partition. The method of claim 13, Means for filtering an environment group from the first partition when the number of statistically significant average ratings is less than or equal to a second threshold, for the environment group of the first partition, Wherein said average rating determines a statistically meaningful user's interest when the number of data items is greater than or equal to a third threshold. The method of claim 13, Means for creating a second partition having an environment group having a number of data items that is greater than or equal to the first threshold; And For each data item in an environment group in the second partition, the data item has the first partition having a minimum Euclidean distance between the center of the cluster in the first partition and the environment portion of the data item. And adding means to the cluster in the system. The method of claim 13, Means for creating a second partition having an environment group having a number of data items that is greater than or equal to the first threshold; And And means for merging all environment groups in the second partition into one outlier cluster. The method of claim 15, And after adding the data item to a cluster in the first partition, means for recalculating centroids for each cluster in the first partition. The method of claim 13, Means for receiving a new data item; Means for placing the new data item in a cluster in the first partition; And Means for dividing the cluster in which the new data item is located when the cluster is about to have two clusters having a number of data items above the first threshold. The method of claim 13, Wherein each environment portion has a plurality of environment values, each environment value determining interest of a user representing a description of a property associated with a state including a usage pattern. A data item associated with a usage pattern of the user, the data item including an interest portion having at least one rating indicating a degree of interest the user has for at least one interest and an environment portion having information about the environment associated with the rating; Storing; Grouping the data items into an environment group having a data item and an associated environment portion; For each environment group, determining whether the number of data items in the environment group is greater than or equal to a first threshold value; Creating a first partition having an environment group having a number of data items above the first threshold value; Averaging ratings for interests of data items belonging to an environment group in the first partition and clustering each environment group in the first partition; And And comparing a current environment with an environment group in the first partition to obtain user interest.

Images