CN102075851A

CN102075851A - Method and system for acquiring user preference in mobile network

Info

Publication number: CN102075851A
Application number: CN2009102385046A
Authority: CN
Inventors: 孟祥武; 张玉洁; 王立才; 张向阳; 王洪明; 张建成
Original assignee: Beijing University of Posts and Telecommunications
Current assignee: Beijing University of Posts and Telecommunications
Priority date: 2009-11-20
Filing date: 2009-11-20
Publication date: 2011-05-25
Anticipated expiration: 2029-11-20
Also published as: CN102075851B

Abstract

The invention discloses a method and system for acquiring user preference in a mobile network. The method comprises the following steps: establishing a user preference acquisition model based on context computation so as to divide available context information into a current context and a historical context; dividing the context computation into current context-aware computation and historical context computation; taking user historical behavior and context thereof as main data sources, and accurately mining the user preference and change thereof step by step by utilizing methods such as collaborative filtering, historical context computation and the like; designing a user preference acquisition system in the mobile network according to the user preference acquisition model by combination with the actual situation of the mobile network so as to inspect the fitting effect of a user preference extraction model in the actual scene of the mobile network and further verify the availability and advanced property of the user preference extraction model; and finally achieving the purposes of extracting group user preference and individual user preference based on different objects from massive information.

Description

The acquisition methods of user preference and system among a kind of mobile network

Technical field

The present invention relates to the business support technology of moving communicating field, relate in particular to the acquisition methods and the system of user preference among a kind of mobile network.

Background technology

Along with 3G (3G (Third Generation) Moblie), 4G (the 4th third-generation mobile communication) network architecture is towards syncretization, the complanation direction develops, mobile communications network is on the basis of merging gradually with computer network, traditional internet information service is extended, for providing more colourful mobile network's service, the user is not only simple communication service, as: traditional voice and data service, the mobile search service, Location based service, the wireless class business of browsing, the mobile flow medium service, the mobile electronic payment service, mobile download service, mobile network game etc., mobile network's service that various Mobile Network Operator and service provider provide is in content, price, QoS (Quality of Service, service quality) etc. also exists than big-difference.Because intelligent mobile device is universal day by day, obtaining and pushing of information resources can occur in " any time, any place, any way ", and the mobile subscriber will be in complicated communication network environment and abundant service provides environment.Yet, under situation limited in one's ability such as the interface display of mobile device, terminal processes, input and output, mobile network's service data of magnanimity, dynamic change is often brought very heavy information burden for the user, thereby will cause " mobile message overload " problem.

Thereby, mobile network's services selection problem, promptly, just seem particularly important and urgent, thereby the key issue that has risen to the solution of mobile network's service field urgent need is paid close attention to more and more how accurately for the mobile subscriber provides its real interested mobile network's COS and content.For example, some research institutions begin to pay close attention to the personalization research of next generation network service field both at home and abroad, and enterprises such as China Mobile, Google company also begin to launch the development of mobile personalized service product.If this problem can solve smoothly, not only can improve mobile network's service experience of user, also can play great facilitation for the good service that Mobile Network Operator and service provider maintain customer group, realization " people-oriented ".

Mobile Network Operator and service provider only after fully, accurately understanding the demand and variation thereof of user to various services, just can provide the service and the content that satisfy its demand; Therefore, making up accurate user preference and obtain model to obtain user's individual demand, is the most key technology.But because the complexity of mobile communications network: on the one hand, mobile network's service is in the environment of height dynamic change, relate to the development of infrastructure network, operation maintenance management, service logic design, content production and distribution, terminal equipment etc., its application is extensive, closely related with the feature of each application industry, and the propelling movement of mobile network service and content thereof can occur in " any time, any place, by any way ", makes that the complexity of research work is very high; On the other hand, the mobile subscriber has different Demographics background and context environmental, the degree of awareness to various services is progressively ripe, and the demand and the preference of serving also just is not quite similar and increasingly sophisticatedization, and this has increased the difficulty of research work undoubtedly again.What therefore, user preference obtained technology among the mobile network is worth research and challenging work.

At the end of last century, along with the fast development of Internet technology and the appearance of user's " information overload " and " information is got lost " problem, people propose to provide for different user the notion of " personalized service " of Differentiated Services.The personalized service research of tradition internet arena is obtained the interest level of user to project to make up the user preference model, mainly based on the two-dimensional space of user-project (user-item), the technology such as collaborative filtering, content-based filtration, hybrid filtration that depend on realize.Although the personalization research of traditional internet arena has obtained many achievements, great majority are soft at desktop, hardware system " information overload " problem.

At present, along with the develop rapidly and the user of mobile communications network are more and more higher to the individual demand of the service of moving, the researcher begins studying aspect the mobile personalized service both at home and abroad, and mobile subscriber's preferential learning also begins to obtain to pay close attention to as the key technology of mobile personalized service.For example, document " G.Lee; S.Bauer; P.Faratin; J.Wroclawski.Learning User Preferences for Wireless ServicesProvisioning.2004:P480-487 " proposes a kind of method of utilizing intensified learning algorithm and Markov model study user preference at dynamic radio service selection problem; Document " Sheng; Q.; B.Benatallah; and Z.Maamar; User-centric services provisioning in wireless environments.Communication of the ACM; Nov, 2008.51 (14): P130-135 " has been described a kind of mechanism that service content is provided towards different user in wireless network environment; Document " H.J.Lee; S.J.Park; MONERS:A News Recommender for the Mobile Web.Expert Systems with Applications; 2007; 32 (1): P143-150 " has then proposed a kind of news commending system towards mobile Internet, the news of recommending by the importance of theme of news self and ageing, user preference changes, the user decides for the preference of classification under the news.Eighties of last century beginning of the nineties, Weiser has proposed the notion of " general fit calculation ", calculates as the context-aware in one of its sub-field of core to begin fully to be paid close attention to.Here, the research contents calculated of described context-aware mainly comprises: context obtains, context modeling and expression, contextual effective utilization, how to make up the system framework of supporting context-aware etc.Its target is to serve and computational resource for the user provides before making system can find and utilize contextual informations such as position, surrounding environment automatically with taking the photograph, thereby reduces man-machine interaction, improves user experience.Described context-aware is calculated and combined with information network is recent years to occur a kind of new technology trends.In next generation network (NGN) environment that merges, the context-aware theory is used for physical layer or network layer (as wireless sensor network) more, application at operation layer (particularly towards mobile subscriber individual business logic) is then less relatively, for example: document " Cheng Bo; MengXiangwu; Chen Junliang.An Adaptive User Requirements Elicitation Framework.IEEE Computer Society Washington; DC; USA; 2007:P501-502 " is considered the effect that context-aware calculating is extracted user's request, proposes the user's request extraction framework that a kind of body drives; Document " Liu Dong; Meng Xiangwu; Chen Junliang.A Framework for Context-AwareService Recommendation.Advanced Communication Technology, ICACT 2008:P2131-2134 " proposes a kind of context-aware service recommendation framework.

But in mobile network's service research field, because the user is usually relevant with various contextual informations to the demand or the preference of mobile network's service and content thereof, and the context environmental than traditional internet arena is complicated more, therefore, the user preference that legacy user's preference pattern not exclusively is fit under mobile network's service environment obtains, and traditional commending system framework also is not suitable for mobile network's services selection model modeling.In a word, in above-mentioned research, context-aware is calculated and to be used to recommend production process mostly, and the less user preference acquisition process that is used for, and also lacks the user preference that calculates based on context among the mobile network is obtained further investigation with the services selection model.

As seen, though it is comparatively deep for Study on Technology such as personalized service, context-aware calculating, user preference obtain both at home and abroad, study also than horn of plenty for UNE service of future generation, service-oriented calculating scheduling theory, still be in the exploratory stage but obtain Study on Technology, demand further further investigation urgently for user preference among the mobile network.

Summary of the invention

In view of this, main purpose of the present invention is to provide the acquisition methods and the system of user preference among a kind of mobile network, obtain user preference according to user's use habit and in conjunction with the context-aware theory of computation, provide personalized service to the user, to improve mobile subscriber's network service experience, simultaneously, can adopt technological means to filter a large amount of redundant informations, make full use of Internet resources, improve Virtual network operator and service provider's service quality and further cut operating costs.

For achieving the above object, technical scheme of the present invention is achieved in that

The system that obtains of user preference among a kind of mobile network, the described system that obtains comprises that user's historical behavior and context generate subsystem (21), storage and ADMINISTRATION SUBSYSTEM (22), data mining subsystem (23) and user preference and extract subsystem (24); Wherein:

User's historical behavior and context generate subsystem (21), generate in order to finish the contextual data of mobile subscriber's historical behavior and user's historical behavior;

Storage and ADMINISTRATION SUBSYSTEM (22) are in order to finish the storage and the management of user's historical behavior data, user's historical behavior contextual information, user preference information;

Data mining subsystem (23) calculates in order to finish user clustering and user's historical behavior context;

User preference extracts subsystem (24), is used for the result of calculation according to described data mining subsystem (23), extracts group user preference and individual consumer's preference information, and outputs to described storage and ADMINISTRATION SUBSYSTEM (22).

Wherein, the described system that obtains further comprises:

User preference adaptive sub system (25) in order to finishing user preference variation/collision detection, user preference correction, and is kept at testing result or correction result in storage and the ADMINISTRATION SUBSYSTEM (22).

Described user's historical behavior and context generate subsystem (21) and further comprise: user's historical behavior generation module (211) and user's historical behavior context generation module (212); Wherein,

Described user's historical behavior generation module (211) is used to realize the data systematic function of user's historical behavior, and its output result is the data source of group user preference extraction;

Described user's historical behavior context generation module (212) is used to realize the contextual data systematic function of user's historical behavior, and its output result is the data source that user's historical behavior context calculates.

Described data mining subsystem (23) further comprises user clustering module (231) and user's historical behavior context computing module (232); Wherein,

Described user clustering module (231), based on the use amount of user to the service of moving, by the utilization clustering algorithm, all users are divided in a plurality of different clusters, make that the user's similarity in the same cluster is higher, user's similarity in the different clusters is lower, and after cluster analysis finished, each user had a cluster labelled notation;

Described user's historical behavior context computing module (232) is used for user's historical behavior context is calculated, in the hope of going out the individual consumer to contextual interest-degree in a certain respect.

Described user preference extracts subsystem (24), further comprises group user preference extraction module (241) and individual consumer's preference extraction module (242); Wherein,

Described group user preference extraction module (241) is used for group user cluster result and group user historical behavior are calculated, output group user preference information;

Described individual consumer's preference extraction module (242) is used for group user preference and user's historical behavior context result of calculation are merged calculating, output individual consumer preference information.

The acquisition methods of user preference among a kind of mobile network, this method comprises:

A, utilize user's historical behavior and context to generate subsystem (21) to generate user's historical behavior data and user's historical behavior context data, and described data are kept in storage and the ADMINISTRATION SUBSYSTEM (22);

B, from described storage and ADMINISTRATION SUBSYSTEM (22), obtain user's historical behavior and context data thereof, and carry out user clustering respectively and the user behavior historical context is calculated by data mining subsystem (23);

C, extract subsystem (24) by user preference again and extract the group user preference according to the user clustering result, and then group user preference and user's historical behavior context result of calculation merged calculating, extracting individual consumer's preference, and user preference information is stored in described storage and the ADMINISTRATION SUBSYSTEM (22).

Further comprise after the described step C:

User preference adaptive sub system (25) is according to user feedback, detect the variation of user's historical behavior and context data thereof, perhaps detect and extracted conflicting of user preference and real user's request, then the user preference that has extracted is carried out the self adaptation correction, and will present to the user by mobile network's service and content thereof through revised output result.

Individual consumer's interest-degree computational methods of calculating based on context among the described mobile network of a kind of claim 1, this method comprises:

The case data information of A, extraction user historical behavior context computing module reads user's historical behavior context database, reads the contextual information of having handled;

B, mate, check whether case exists according to user's historical behavior contextual information of being read inquiry case database, if case exists, execution in step C then; Otherwise, execution in step D;

C, the corresponding case in the case library is made amendment, then execution in step E;

D, the modification that will make case or the new case of creating deposit in the case database, then execution in step E;

E, judge whether that contextual information has read and finish,, then return execution in step A and continue to read contextual information that read until described contextual information and finish, case library is created and finished if do not read; Otherwise, execution in step F;

F, end case are extracted flow process, beginning context calculation process, read case information in the case library, by calculating the precondition value in the Bayesian network, and the condition value that has obtained suddenly according to previous step, calculate the value of Bayesian network of each subnet of separation, draw user's daily behavior custom, the value of each subnet that will draw deposits in the database then;

G, calculate user's service probability of use value according to the value of each precondition and each subnet, use clustering algorithm that user's service probability of use value is carried out cluster analysis, probable value is divided into different grades, extracting individual consumer's interest-degree, and deposit the individual consumer's interest-degree that is extracted in individual consumer's interest-degree database.

Merge the user preference extracting method that collaborative filtering and context calculate among the described mobile network of a kind of claim 1, this method comprises:

A, obtain individual consumer's interest-degree and colony's user preference, travel through each user interest-degree to every class service under every kind of context environmental; B, judge whether the user has traveled through, if do not traveled through, execution in step c then; Otherwise, finish this ergodic process; Whether c, judgement are finished in the traversal of certain class service at certain body and function family, then return step b if finish; Otherwise, execution in step d; Whether d, judgement are finished at the context traversal of described individual consumer under certain class service, then return step c if finish; Otherwise, execution in step e; E, judge that whether described individual consumer's interest-degree is zero, if, execution in step f then; Otherwise, execution in step i; F, travel through this individual consumer place group every other user under this kind context environmental to the interest-degree of such service, and obtain this user place group identification, execution in step g then; G, calculate described every other user effective mean value to the interest-degree of such service under this kind context environmental; H, judge that whether described effective mean value be zero, if, execution in step i then; Otherwise, execution in step j; I, individual consumer's preference value are the preference value of this user place group to such service, execution in step l then; J, individual consumer's preference value are effective mean value; Execution in step l then; K, individual consumer's preference value are this individual consumer's interest-degree, then execution in step l; L, individual consumer's preference is stored in the corresponding database, return execution in step d then.

The acquisition methods of user preference and system among the mobile network provided by the present invention have the following advantages:

User preference obtains model among a kind of mobile network who calculates based on context by proposing in the present invention, and design the validity that corresponding prototype system realizes obtaining with this user preference model, wherein, by using user's historical behavior context generation module, the residing context environmental of user when the recording user historical behavior takes place, and these historical context are calculated, thereby solve the user preference that calculates based on context, will help improving the accuracy of personalized service; In user's historical behavior context computing module, adopt the user interest degree computational methods of calculating based on context, historical context is carried out reasoning and utilized bayesian theory to calculate user's historical behavior context, various context probability when solving the generation of user's historical behavior, and finally utilize individual consumer's interest-degree of the formal description of various dimensions matrix, enriched the precision of user preference and described; Extract the user preference extracting method that adopts fusion collaborative filtering technology and context to calculate in the subsystem at user preference, the individual consumer's interest-degree that will calculate based on context and the service preferences of individual consumer place group merge calculating, thereby prediction and extraction different user to the preference of different mobile networks' services, finally produce the required user preference information with the description of various dimensions matrix form under different context environmentals.In addition, in user preference adaptive sub system, implemented user preference adaptive approach based on context calculating, detect by variation and conflict, and the result of user preference information is revised, further improved the accuracy of extracting user preference information user preference.

Description of drawings

Fig. 1 obtains the model schematic diagram for the user preference that calculates based on context in the embodiment of the invention;

Fig. 2 obtains the system configuration schematic diagram for user preference among the mobile network of the present invention;

Fig. 2 A generates user's historical behavior generation module illustrative view of functional configuration of subsystem for user's historical behavior among Fig. 2 of the present invention and context;

Fig. 2 B generates user's historical behavior context generation module illustrative view of functional configuration of subsystem for user's historical behavior among Fig. 2 of the present invention and context;

Fig. 2 C is the user clustering functions of modules structural representation of data mining subsystem among Fig. 2 of the present invention;

Fig. 2 D is user's historical behavior context computing module illustrative view of functional configuration of data mining subsystem among Fig. 2 of the present invention;

Fig. 2 E extracts the group user preference extraction module illustrative view of functional configuration of subsystem for user preference among Fig. 2 of the present invention;

Fig. 2 F extracts individual consumer's preference extraction module illustrative view of functional configuration of subsystem for user preference among Fig. 2 of the present invention;

Fig. 2 G is the illustrative view of functional configuration of user preference adaptive sub system among Fig. 2 of the present invention;

Fig. 3 obtains the physical structure schematic diagram of system for mobile network user preference in the embodiment of the invention;

Fig. 3 A is the basic handling schematic flow sheet that user preference obtains system among the mobile network of the present invention;

Fig. 4 is historical behavior context generative process schematic diagram in the prototype system of the present invention;

Fig. 5 is the schematic flow sheet of individual consumer's interest-degree computational methods of calculating based on context in the prototype system of the present invention;

Fig. 6 is the user preference extracting method schematic flow sheet of fusion collaborative filtering in the prototype system of the present invention and context calculating.

Embodiment

Below in conjunction with accompanying drawing and embodiments of the invention method of the present invention is described in further detail.

Core concept of the present invention is: obtain model by setting up a user preference that calculates based on context, on the basis of traditional individuation service system, take into full account the importance of contextual information in the mobile network user preference is obtained of complicated dynamic change, available contextual information is divided into current context and historical context, and therefore contextual calculating is divided into current context perception calculating and historical context and calculates, then with user's historical behavior and context thereof as main data source, by progressively accurately digging user preference and the variation thereof of methods such as collaborative filtering and historical context calculating; Obtain model according to described user preference then, actual conditions in conjunction with mobile phone users, Mobile Network Operator and Internet Service Provider, design that user preference obtains system among the mobile network, checking described user preference extraction model, thereby verify the validity and the advance of described user preference extraction model to the fitting effect in mobile network's actual scene.Described prototype system essence is an independently system, its function is cooperatively interacted, is finished jointly by some subsystems and module, its function is corresponding with the function that described model should be finished, and prototype system mainly comprises: user's historical behavior and context generate subsystem, storage and ADMINISTRATION SUBSYSTEM, data mining subsystem, user preference extraction subsystem, user preference adaptive sub system and user interface.Generate subsystem, storage and ADMINISTRATION SUBSYSTEM, data mining subsystem, user preference adaptive sub system and user preference extraction subsystem by user's historical behavior and the context that utilizes described prototype system, finally realize from magnanimity information, extracting at the group user preference of different objects and the purpose of individual consumer's preference.

Fig. 1 obtains the model schematic diagram for the user preference that calculates based on context in the embodiment of the invention, as shown in Figure 1, this user preference obtains model and belongs to the data driven type model, it can be divided into data collection layer from low to high from functional plane, user preference securing layer and user preference adaptation layer, this model uses user preference extractive technique and user preference adaptive technique, the object of its research, it is data message, mainly from mobile network's service provider, described data message comprises: user's historical behavior information, user's historical behavior contextual information, user's demographic information, mobile network's information on services, contextual information etc.; Also can be divided into three classes: i.e. user profile, resource information and contextual information by the source of data.

The mathematical description that the user preference that calculates based on context obtains model tentatively is: M={U, I, C, P}, U * I * C → P.Wherein: U representative of consumer information, I represents the object resource information, and C represents contextual information, P representative of consumer preference.Model Calculation is according to mainly being user's historical behavior and user's historical behavior context, obtained by " data collection layer " of model bottom.Wherein, user's historical behavior is used to describe the operating position of user to object resource (current model is an object with mobile network's service); User's historical behavior context is used to describe the user residing context condition when using the object resource.

Because the extraction of user preference is a progressively accurate process, user in the model shown in Figure 1 " preference extract layer " at first will extract comparatively coarse group user preference by calculating user's historical behavior, extract individual consumer's interest-degree by calculating user's historical behavior context then, again two kinds of data fusion are calculated at last, extracted comparatively accurate individual consumer's preference.

Here, the user preference self adaptation, be meant detect that user preference changes or the conflicting of the user preference that extracts and real user's request, and the technology of revising.The user preference adaptive technique is to realize that by described " user preference adaptation layer " it mainly comprises two necessary processes: the first, user preference variation/collision detection; The second, user preference correction.

Because it is the basis of realizing that user preference obtains system among the described mobile network that user preference of the present invention obtains the structure of model, therefore need obtain theory or the technology that uses in the model to described user preference, as: body, collaborative filtering, context calculating, matrix theory and Markov process etc. are done one and are simply introduced:

Described body, the clear and definite formalization normalized illustration of shared ideas model.Based on body user, object resource, context, user preference etc. are carried out modeling, help from semantic level above-mentioned every field data are carried out structuring, formalized description, be convenient to inferring the complex relationship of every field data.

Described collaborative filtering, collaborative filtering method utilize the similitude (the perhaps similitude between the project) between the user, provide decision support based on other people evaluation and recommendations for the user.Be mainly used in the process that the group user preference is extracted among the present invention.

Described context calculates, and refers to a kind of computation schema that system can find and effectively utilize contextual information (as equipment and personnel, the User Activity etc. of customer location, time, environmental parameter, vicinity) to calculate.The model that proposes in the invention is divided into two kinds with context, i.e. historical context, current context; Thereby context calculating also is divided into two parts, i.e. historical context calculating, current context perception calculating.When obtaining user preference in the present invention, the residing historical context of user when considering the generation of user's historical behavior; When producing recommendation, consider the current context of user.

Described matrix theory, user preference obtains model and investigates user preference from multidimensional angles such as user, service, contexts, thereby use the multi-dimensional matrix theory that each the denapon element that influences user preference is calculated, and the user preference that gets access to is stored in the multi-dimensional matrix.

Described markoff process is meant event at a time, and just the random process that exerts an influence in the limited period belongs to theory of random processes.The markoff process theory is applied in the user preference self adaptation calculating process, can effectively detects and the compute user preferences variation, thereby carry out the user preference correction.

Fig. 2 obtains the system configuration schematic diagram for user preference among the mobile network of the present invention, as shown in Figure 2, described prototype system mainly is made of user's historical behavior and context generation subsystem 21, storage and ADMINISTRATION SUBSYSTEM 22, data mining subsystem 23, user preference extraction subsystem 24 and user preference adaptive sub system 25.Wherein,

User's historical behavior and context generate subsystem 21, generate in order to finish the contextual data of mobile subscriber's historical behavior and user's historical behavior.

Here, described user's historical behavior and context generation subsystem 21 also comprises: user's historical behavior generation module 211, user's historical behavior context generation module 212.

Described user's historical behavior generation module 211 is used to realize the data systematic function of user's historical behavior, and its output result can be used as the data source that the group user preference is extracted.

Described user's historical behavior context generation module 212 is used to realize the contextual data systematic function of user's historical behavior, and its output result can be used as the data source that user's historical behavior context calculates.

Need to prove, only select those to influence the most deep context type of user preference in the present invention, as time, position, use equipment and User Activity state etc., but this does not limit the described user preference that calculates based on context and obtains the context that model is only supported above-mentioned several types, described model does not also rely on concrete context type, can also further expand, as the social environment that increases weather, noisy environment, light, user etc.

Storage and ADMINISTRATION SUBSYSTEM 22 are in order to finish the storage and the management of user's historical behavior data, user's historical behavior contextual information, user preference information.

Here, described storage and ADMINISTRATION SUBSYSTEM 22 further comprise user preference storage and management module 221, user's historical behavior storage and management module 222 and user's historical behavior context storage and management module 223.Wherein: described user preference storage and management module 221 is used for storage and leading subscriber preference information.Described user's historical behavior storage and management module 222 is used for storage and leading subscriber historical behavior information.Described user's historical behavior context storage and management module 223 is used for storage and leading subscriber historical behavior contextual information.

Described storage and ADMINISTRATION SUBSYSTEM 22, the tables of data relevant by user message table, user's historical behavior table, user's historical behavior contextual information table, group user preference table, individual consumer's preference table and user interface manages Various types of data information.Storage and ADMINISTRATION SUBSYSTEM 22 mainly manages data message user message table, user's historical behavior table, user's historical behavior contextual information table, group user preference table, individual consumer's preference table and the data structure that is associated thereof.

Data mining subsystem 23 calculates in order to finish user clustering and user's historical behavior context.

Here, described data mining subsystem 23 further comprises user clustering module 231 and user's historical behavior context computing module 232; Wherein,

Described user clustering module 231, based on the use amount of user to the service of moving, by utilization K-Means clustering algorithm, all users are divided in K the different cluster, make that the user's similarity in the same cluster is higher, user's similarity in the different clusters is lower, and after cluster analysis finished, each user had a cluster labelled notation.

User's historical behavior context computing module 232 is used for user's historical behavior context is calculated, in the hope of going out the individual consumer in a certain respect interest-degree.

User preference extracts subsystem 24, is used for the result of calculation according to described data mining subsystem 23, extracts group user preference and individual consumer's preference information and output from described storage and ADMINISTRATION SUBSYSTEM 22; Also be used for extracting and export revised group user preference and individual consumer's preference information according to the detection/correction result of described user preference adaptive sub system 25 and in conjunction with the result of calculation of described data mining subsystem 23.

Here, described user preference extracts subsystem 24, further comprises group user preference extraction module 241 and individual consumer's preference extraction module 242; Wherein,

Described group user preference extraction module 241 is used for group user cluster result and group user historical behavior are calculated, output group user preference information.

Described individual consumer's preference extraction module 242 is used for group user preference and user's historical behavior context result of calculation are merged calculating, output individual consumer preference information.

User preference adaptive sub system 25 in order to finishing user preference variation/collision detection, user preference correction, and is kept at testing result or correction result in storage and the ADMINISTRATION SUBSYSTEM 22.

Here, user preference adaptive sub system 25 further comprises user preference variation/collision detection module 251 and user preference correcting module 252.Wherein:

Described user preference variation/collision detection module 251 is used for the variation of in time catching user preference according to the change threshold or the change detection algorithm of setting in advance; Or be used to detect when the user preference of multi-source channel extraction merges, it is inconsistent and produce the phenomenon of conflict user preference to occur, or detect and to exist inconsistent and phenomenon that generation conflicts between the preference of institute's user preference that extracts and the explicit setting of user, and described user preference variation/conflict situations is passed to user preference correcting module 252.

Described user preference correcting module 252, be used to finish the debugging functions of user preference, i.e. different objective circumstances that change at user preference conflict and user preference, set the user preference correction algorithm, to reach user preference information is revised, further the effect of the feasible user preference information precision that is obtained.

In addition, outside the prototype system, also comprise the user interface of information interaction between a responsible user and the described prototype system.Here, described user comprises mobile phone users and system management maintenance customer.

Fig. 2 A is user's historical behavior generation module illustrative view of functional configuration of user's historical behavior among Fig. 2 of the present invention and context generation subsystem, and shown in Fig. 2 A, user's historical behavior generation module 211 further comprises following submodule:

Read external information document submodule 2111, be used for to external world information document and resolve, and analysis result deposited in specify in the array for other submodules uses.

Initialization user profile submodule 2112 is used to generate user ID, it is deposited in the database, and give each attribute field assignment of this user.

Initialization user behavior table submodule 2113, the user ID (major key) that is used for according to user message table is come initialization user historical behavior table, and described user's historical behavior table comprises major key (user ID) and each volume of services of each bar record in the user behavior table.

Generate user behavior data submodule 2114, the analysis result array that is used for reading external information document submodule 2111 is as input, and with the Data Update that the generates respective field in the database user behavior table.

User's historical behavior data submodule 2115 is used to realize the increase and decrease of user's historical behavior data, increases and decreases on the basis as in the month before historical data, and it is saved in the database.

DB Backup submodule 2116 is used for the backup of fulfillment database information, so that check the data that produce early stage.

Fig. 2 B is user's historical behavior context generation module illustrative view of functional configuration of user's historical behavior among Fig. 2 of the present invention and context generation subsystem, and shown in Fig. 2 B, user's historical behavior context generation module 212 further comprises following submodule:

User's historical behavior context generates parent reason submodule 2121, be used to start the parent reason, the promoter agency, receive the control messages of superior agency, and from database, obtain user's volume of services data, generate inlet module as whole user's historical behavior contextual information, this submodule is controlled the operating procedure and the process of other module, the agency of this submodule is the operation that self each module is controlled in the control information of user's historical behavior generation module by receiving higher level's submodule, after the control messages that receives higher level's submodule, this submodule agency begins to start other each agency, and passes through the operation rhythm of each proxy module of user profile data flow con-trol.

User profile generates acts on behalf of submodule 2122, be used to generate user's user profile, the user profile that generates is stored in the user message table, act on behalf of submodule by this, the mechanics that can meet oneself for user's information setting according to different situations, after this acts on behalf of the submodule operation, can generate the user profile that meets certain rule for the user according to the rule that the user sets.

Facility information generates acts on behalf of submodule 2123, be used to each service to generate subscriber equipment, make the corresponding a kind of subscriber equipment of each bar service recorder, and deposit the contextual information that generates in the user context information table, with above-mentioned module, this submodule also can be provided with different rules, allow the user of different characteristic use different equipment, because it is separate respectively acting on behalf of between the submodule, therefore when different rules is set, the current agency's that other agency seldom can influence logic, and this change of acting on behalf of internal logic can not have influence on the realization that other act on behalf of logic fully.

Temporal information generates acts on behalf of submodule 2124, be used to each service to generate user's time context, make each corresponding different time of bar record, with the above-mentioned described submodule of acting on behalf of, this acts on behalf of submodule also can generate the logic of oneself according to the rule of the self-demand of setting, the modification of this agency's logic does not influence other agency, in the historical behavior context database table that deposits relative users in that bears results.

Positional information generates and acts on behalf of submodule 2125: be used to each professional context that generates user locations, make the corresponding user position of each bar service recorder; This agency's formation logic can be fully according to different users, and different identity informations produces different positional informations, and the contextual location information of Sheng Chenging can provide better foundation for work from now on the contextual information in the very approaching reality like this.

Action message generates acts on behalf of submodule 2126: the action message context for each service generation user historical behavior makes all corresponding a kind of User Status of each bar service recorder.Act on behalf of submodule in like manner with above-mentioned, according to different user profile, generate different user activity information, the rule that action message produces is provided with in the described submodule inside of acting on behalf of, according to different environments for use different rules is set, thus the final contextual information record that meets user's request that produces.

Volume of services assignment agent submodule 2127: the volume of services of user's historical behavior is distributed, allow the volume of services of various services well be assigned to each section in one month 30 days in the period, the distribution of volume of services can make and at random total amount is assigned in all days, also can be according to must use habit distribution services amount, service logic among the agency can change at any time in a word, variation by different service logics can adapt to different environment, satisfies various demands.

Fig. 2 C is the user clustering functions of modules structural representation of data mining subsystem among Fig. 2 of the present invention, and shown in Fig. 2 C, user clustering module 231 further comprises following submodule:

Extract user profile submodule 2311, be used for extracting user's ID number and user's historical behavior information, i.e. Fu Wu use amount from depositing user's historical behavior database of information.

Data map submodule 2312 is used for all data are mapped on the same interval [a, b] according to certain reflection method, and purpose is in order to reduce influencing each other between the multidimensional data attribute, avoids big number to flood decimal, thereby makes the cluster effect desirable more.

Seek cluster centre submodule 2313, be used to produce K mutually different random number and determine initial cluster center, and select initial cluster center.

Calculate cluster centre submodule 2314, be used to calculate the cluster centre of each cluster; The method of calculating cluster centre be mean value with all data in each cluster as cluster centre, specific practice is: multidimensional data calculates the mean value of every dimension.

Cluster is divided submodule 2315, is used for all data are divided into nearest with it cluster, and division methods is the Euclidean distance of calculated data and each cluster centre.

Algorithmic statement is judged submodule 2316, is used for judging that by the condition of convergence whether the cluster division finishes, if satisfy the condition of convergence, then finishes clustering algorithm; Otherwise, then proceed cluster and divide, till satisfying the condition of convergence.

User clustering number is provided with submodule 2317, is used to each user in the database that a poly-class-mark is set.

Fig. 2 D is user's historical behavior context computing module illustrative view of functional configuration of data mining subsystem among Fig. 2 of the present invention, and shown in Fig. 2 D, user's historical behavior context computing module 232 further comprises following submodule:

＜the time | the position | activity, move service〉case extraction submodule 2321, this module is by obtaining the context service recorder information that has generated, by extracting time, position, activity, the information on services foundation＜time wherein | the position | activity, move service〉case library, in this case leaching process, read the context record information of generation one by one, set up case, removed, only keep single case for the case that repeats.Each user is had corresponding case library so that next step is found the solution.

＜the time, the position〉daily habits case extraction submodule 2322, this module generates context service recorder information by obtaining, by extraction time, positional information foundation＜time, position〉case library.In this case leaching process, after definite demand, read the context record information of generation one by one to the case type, set up case, removed for the case that repeats, only keep single case.Each user is had corresponding case library so that next step is found the solution.

＜position, activity〉the daily habits case extracts submodule 2323, this module generates context service recorder information by obtaining, by extracting position, activity contexts information foundation＜position, activity〉case library.In this case leaching process, after definite demand, read the context record information of generation one by one to the case type, set up case, removed for the case that repeats, only keep single case.Each user is had corresponding case library so that next step is found the solution.

＜the time | the position | activity, move service〉interest-degree calculating sub module 2324, according to＜time | the position | activity, the service of moving〉case extracted of interest-degree calculating sub module, calculate the interest-degree of user for corresponding case, because the value of case correspondence is different (as the use amount differences of the service of moving) in leaching process, different case is calculated also difference of gained interest level at last.By using Bayesian formula the calculating of posterior probability can be changed into the calculating of prior probability, make good solution of calculating of interest-degree and realization.Bayesian network organically combines directed acyclic graph and probability theory, causality is showed intuitively with directed graph, and the statistical value of historical record all is discrete data, makes Bayesian formula be applied to have good effect aspect the statistical computation of mobile field historical record.

＜the time, the position〉daily habits interest-degree calculating sub module 2325, according to＜time, the position〉case extracted of daily habits interest-degree calculating sub module, calculate the interest-degree of user for corresponding case, because case time corresponding value is different in leaching process, so different case is calculated also difference of gained interest level at last.By Bayesian formula the calculating of posterior probability is changed into the calculating of prior probability, make good solution of calculating of interest-degree and realization.

＜position, movable〉daily habits interest-degree calculating sub module 2326, according to＜position, movable〉case extracted of daily habits interest-degree calculating sub module, calculate the interest-degree of user for corresponding case, because the value of case correspondence is different in leaching process, so different case is calculated also difference of gained interest level at last.The calculating of posterior probability can be changed into the calculating of prior probability again by Bayesian formula, can calculate corresponding interest level simply efficiently.

Fig. 2 E is the group user preference extraction module illustrative view of functional configuration of user preference extraction subsystem among Fig. 2 of the present invention, and shown in Fig. 2 E, group user preference extraction module 241 further comprises following submodule:

Group's historical behavior matrix construction submodule 2411, be used to finish the function of group's historical behavior matrix construction, be input promptly, calculate the average use amount of all users of each group, structure group historical behavior matrix all kinds of mobile network's services with the user clustering result.

Service cluster submodule 2412 is used to finish the function of serving cluster, is input with group's historical behavior matrix promptly, historical behavior data with each group are object to be clustered, and according to clustering algorithm, each group's operating position cluster that every class is served is 5 (or 3) grades, that is: 5-preference very, 4-is than preference, the 3-general preference, and 2-is preference more not, 1-very not preference (or, the senior preference of 3-, the 2-general preference, 1-is preference more not).

Group user preference calculating sub module 2413, be used to finish group user preference computing function, be input promptly with group's historical behavior matrix and service cluster result, according to the service cluster result group's historical behavior matrix is mapped as the group user preference matrix, wherein, described matrix adopts the two dimensional form of " group-service ", with group's row vector by name, with all kinds of services is column vector, and the group user preference matrix is stored in the group user preference storage and management module in storage and the ADMINISTRATION SUBSYSTEM.

Group user preference structure descriptor module 2414, be used to finish the structural description function of group user preference, promptly extracting the result with the group user preference is input, the semi-structured group user preference that generates the XML form according to the XML Schema that has designed is described document, and document storage is arrived group user preference storage and management module.

Fig. 2 F is individual consumer's preference extraction module illustrative view of functional configuration of user preference extraction subsystem among Fig. 2 of the present invention, and shown in Fig. 2 F, individual consumer's preference extraction module 242 further comprises following submodule:

Individual consumer's interest-degree zero value detection submodule 2421 is used to finish individual consumer's interest-degree zero value detection function, and promptly whether certain user interest-degree to certain class service under certain specific context condition is zero in individual consumer's interest-degree database; If be zero, then carry out individual consumer's preference calculating sub module function; Otherwise, carry out effectively average interest-degree calculating sub module.

Effectively average interest-degree calculating sub module 2422, be used to finish effectively average interest-degree computing function, promptly effectively average interest-degree calculating sub module need find other users of group under the individual consumer from group user preference matrix database, and the nonzero value interest-degree that these users of traversal serve certain class under certain specific context condition in individual consumer's interest-degree database, and calculate their mean value, be output as effectively average interest-degree.

Effectively average interest-degree zero value detection submodule 2423 is mainly finished effectively average interest-degree zero value detection function, i.e. whether the effectively average interest-degree that the effectively average interest-degree calculating sub module of detection is exported is zero.If be zero, then carry out individual consumer's preference calculating sub module function, and do not visit group's user preference database; Otherwise, carry out individual consumer's preference computing module function, and visit group user preference database.

Individual consumer's preference calculating sub module 2424, be used to finish individual consumer's preference computing function, be specially: according to the result of calculation and the Rule of judgment of above-mentioned module, under different conditions, the value of individual consumer's preference is set to individual consumer's interest-degree, effectively average interest-degree or individual place group user preference respectively, thereby dope this user preference value to the inhomogeneity service under different context environmentals, and will export the result and store in individual consumer's preference database (individual consumer's preference storage and management module).

Individual consumer's preference structure descriptor module 2425, be used to finish the structural description function of individual consumer's preference, promptly extracting the result with individual consumer's preference is input, the semi-structured individual consumer's preference that generates the XML form according to the XML Schema that has designed is described document, and document storage is arrived individual consumer's preference storage and management module.

Fig. 2 G is the illustrative view of functional configuration of user preference adaptive sub system among Fig. 2 of the present invention, shown in Fig. 2 G, described user preference adaptive sub system 25 is by the variation of using user preference variations/collision detection module 251 in time to catch the group/individual user preference according to the change threshold or the change detection algorithm of setting; Or when the user preference that the multi-source channel is extracted merges, in time detect user preference inconsistent and produce conflict or inconsistent and conflicting of causing between the preference of the user preference that extracts and the explicit setting of user; And then utilize described user preference correcting module 252 to revise, and correction result is kept in the user preference database.

Fig. 3 obtains the physical structure schematic diagram of system for mobile network user preference in the embodiment of the invention, there is shown and comprise operator, service content provider, the terminal use is the relation in the described prototype system under interior mobile network environment, described storage and ADMINISTRATION SUBSYSTEM write down Virtual network operator and serve the data message that content supplier uses mobile device to produce for the terminal use, by using user preference to extract subsystem and user preference adaptive sub system, in conjunction with context environmental the data message in described storage and the ADMINISTRATION SUBSYSTEM is handled, in the hope of obtaining described terminal use's group's preference and user preference information.

Fig. 3 A is the basic handling schematic flow sheet that user preference obtains system among the mobile network of the present invention, below in conjunction with Fig. 3 the acquisition methods of mobile network user preference in the prototype system of the present invention is further detailed:

With reference to figure 3A, information interactive process between each subsystem of described prototype system comprises the steps:

Step 301～302: trigger user's historical behavior and context generation subsystem 21 from user interface, generate user's historical behavior data and user's historical behavior context data;

Step 303: described user's historical behavior and context generation subsystem are stored into user's historical behavior and the context thereof that generates in storage and the ADMINISTRATION SUBSYSTEM 22;

Step 304～305: data mining subsystem 23 obtains user's historical behavior and context data thereof from storage and ADMINISTRATION SUBSYSTEM 22, carries out user clustering and user behavior historical context respectively and calculates;

Step 306～313: user preference extracts subsystem 24 and extracts the group user preference according to the user clustering result, and then group user preference and user's historical behavior context result of calculation merged calculating, with extraction individual consumer preference, and store storage and ADMINISTRATION SUBSYSTEM 22 into;

Step 314: select to be fit to the mobile network COS of this user according to user preference, current context perception information, and recommend the user at current context environment needs;

Step 315: user interface is presented to the user with recommended mobile network's service and content thereof, and feeds back;

Step 316～317: user preference adaptive sub system 25 is according to user feedback, detect the variation of user's historical behavior and context data thereof, perhaps detect and extracted conflicting of user preference and real user's request, then the user preference that has extracted is carried out the self adaptation correction, and will promptly the user be presented in recommended mobile network's service and content thereof through revised output result by user interface.

Fig. 4 is historical behavior context generative process schematic diagram in the prototype system of the present invention, user's historical behavior context generation module 212 with the output result of user's historical behavior generation module 211 as input, at first generate user's personal information for the user, it is the unit rise time with mobile network's service then, reproducing device, and then generation position (scope of the customer location of user's personal information decision), next will generate user's activity (the generation result of User Activity depends on user position information), be that user's volume of services distributes at last.More than be not separate in each generative process, also exist between each agency to rely on and restriction relation, and finally generate user's historical behavior contextual information.As shown in Figure 4, this process specifically comprises:

Step 401: start the historical behavior context and generate parent reason and each sub agent, receive the control messages of superior agency (user's historical behavior generates the agency).

Step 402: the historical behavior context generates the parent reason and obtain user's historical behavior data from database.

Step 403: user profile generates the agency and receives the control messages that the historical behavior context generates the parent reason, and the variable of control messages transmission is make sure for the user ID kimonos and used total amount.

Step 404: user profile generates the user profile that the agency generates the user, the user profile that generates is stored in the user message table, and transmit control message to subscriber equipment generation agency.

Step 405: use subscriber equipment to generate the agency and generate subscriber equipment, and transmit control message to time generation agency for each user.

Step 406: generating service time and act on behalf of, is the time context that every class service generates the user, makes each corresponding different time of bar record, and transmits control message to position generation agency.

Step 407: generation agency in place to use is the place context that every class service generates the user, makes the corresponding user position of each bar service recorder, and transmits control message to User Activity generation agency.

Step 408: the use User Activity generates the agency, and the action message context for every class service generation user historical behavior makes all corresponding a kind of User Status of each bar service recorder, and transmits control message to volume of services generation agency.

Step 409: use the volume of services assignment agent, user's historical behavior volume of services to each user distributes, allow the volume of services of various mobile networks service be assigned to each section in one month 30 days in the period, and will be with contextual user's historical behavior to store in user's historical behavior context database.Need to prove, distribute one day/24 hours volume of services at every turn, the subscriber equipment that transmits control message then generates the agency, carries out the circulation of this user's the distribution of volume of services next time.

Step 410: the volume of services assignment agent subscriber equipment that transmits control message generates the agency.

Step 411: after the time generated user behavior that the agency detects this user one month/30 day and assigns, transmitting control message generated the parent reason to the historical behavior context, and user's historical behavior context data of carrying out next user generates.

In addition, in user's historical behavior context data generative process, use the multi-thread concurrent control technology, each circulation generates the individual user's data of N (getting 5～10).

Fig. 5 is the schematic flow sheet of individual consumer's interest-degree computational methods of calculating based on context in the prototype system of the present invention, with user's historical behavior context data as input, by context knowledge being carried out reasoning and utilizing bayesian theory to calculate user's historical behavior context, various context probability when solving the generation of user's historical behavior, final output comprises individual consumer's interest-degree of contextual information, as shown in Figure 5, this process comprises:

Step 501: the case data information of extracting user's historical behavior context computing module;

Step 502: read user's historical behavior context database, read the contextual information of having handled;

Here, described contextual information is the contextual information after the raw information process processing of domain body that transducer obtains, and this moment, contextual information no longer was original numeral, but a kind of semantic information;

Step 503: mate according to the user's historical behavior contextual information inquiry case database that is read, check whether case exists, if case exists, then execution in step 504; Otherwise, execution in step 505;

Step 504: the corresponding case in the case library is made amendment, and execution in step 506 then;

Step 505: create a new case, execution in step 506 then;

Step 506: will modification or the new case of creating that case is made be deposited in the case database, execution in step 507 then;

Step 507: judging whether that contextual information has read finishes, if do not read, then returns execution in step 501 and continues to read contextual information, reads until described contextual information to finish, and case library is created and finished; Otherwise, execution in step 508;

Step 508: finish case extraction flow process and begin the context calculation process, execution in step 509;

Step 509: read case information in the case library, by calculating the precondition value in the Bayesian network, execution in step 510 then;

Step 510: according to the condition value that previous step has been obtained suddenly, calculate the value of Bayesian network of each subnet of separation, promptly draw user's daily behavior custom, execution in step 511 then;

Step 511: the value of each subnet that will draw deposits in the database;

Step 512: the service probability of use value of calculating the user according to the value of each precondition and each subnet;

Step 513: use clustering algorithm that user's service probability of use value is carried out cluster analysis, probable value is divided into different grades, to extract individual consumer's interest-degree;

Step 514: deposit the individual consumer's interest-degree that is extracted in individual consumer's interest-degree database;

Step 515: finish computational process.

Fig. 6 is the user preference extracting method schematic flow sheet of fusion collaborative filtering in the prototype system of the present invention and context calculating, obtain model as can be known with reference to user preference shown in Figure 1, this method is to be input with user's historical behavior and user's historical behavior contextual information, utilization is calculated user's historical behavior based on the collaborative filtering method of clustering algorithm, and extracts the group user preference; Utilize the historical context computational methods to calculate user's historical behavior context, and extract individual consumer's interest-degree; To merge calculating based on the group user preference of collaborative filtering method and the individual consumer's interest-degree that calculates based on context then, thereby prediction and extraction different user preference to different mobile networks' services under different context environmentals, i.e. individual consumer's preference information.With reference to figure 6, this method comprises the steps:

Step 601: obtain individual consumer's interest-degree and colony's user preference, travel through each user interest-degree to every class service under every kind of context environmental;

Step 602: judge whether the user has traveled through, if do not traveled through, then execution in step 603; Otherwise if execute, then execution in step 613;

Step 603: judge at certain body and function family whether finish in the traversal of certain class service, then return step 602 if finish; Otherwise, execution in step 604;

Step 604; Whether judgement is finished at the context traversal of described individual consumer under certain class service, then returns step 603 if finish; Then, execution in step 605;

Step 605: whether the interest-degree of judging described individual consumer is zero, if then execution in step 606; Otherwise, execution in step 611;

Step 606: the every other user who travels through this individual consumer place group to the interest-degree of such service, and obtains this user place group identification under this kind context environmental, execution in step 607 then;

Step 607: calculate described every other user mean value to the interest-degree of such service under this kind context environmental; Mean value described here is the mean value of the interest-degree of nonzero value, promptly effective mean value;

Step 608: judge whether described effective mean value is zero, if then execution in step 609; Then, execution in step 610;

Step 609: individual consumer's preference value is the preference value of this user place group to such service, and execution in step 612 then;

Step 610: individual consumer's preference value is effective mean value; Execution in step 612 then;

Step 611: individual consumer's preference value is this individual consumer's a interest-degree, and execution in step 612 then;

Step 612: individual consumer's preference is stored in the corresponding database, return execution in step 604 then;

Step 613: finish this ergodic process.

The above is preferred embodiment of the present invention only, is not to be used to limit protection scope of the present invention.

Claims

1. the system that obtains of user preference among the mobile network, it is characterized in that the described system that obtains comprises that user's historical behavior and context generate subsystem (21), storage and ADMINISTRATION SUBSYSTEM (22), data mining subsystem (23) and user preference and extract subsystem (24); Wherein:

2. the system that obtains of user preference is characterized in that among the mobile network according to claim 1, and the described system that obtains further comprises:

3. the system that obtains of user preference among the mobile network according to claim 1, it is characterized in that described user's historical behavior and context generate subsystem (21) and further comprise: user's historical behavior generation module (211) and user's historical behavior context generation module (212); Wherein,

4. the system that obtains of user preference is characterized in that among the mobile network according to claim 1, and described data mining subsystem (23) further comprises user clustering module (231) and user's historical behavior context computing module (232); Wherein,

5. the system that obtains of user preference is characterized in that among the mobile network according to claim 1, and described user preference extracts subsystem (24), further comprises group user preference extraction module (241) and individual consumer's preference extraction module (242); Wherein,

6. the acquisition methods of user preference among the mobile network is characterized in that this method comprises:

7. the acquisition methods of user preference is characterized in that among the mobile network according to claim 6, further comprises after the described step C:

8. individual consumer's interest-degree computational methods of calculating based on context among the described mobile network of claim 1 is characterized in that this method comprises:

9. merge the user preference extracting method that collaborative filtering and context calculate among the described mobile network of claim 1, it is characterized in that this method comprises:

A, obtain individual consumer's interest-degree and colony's user preference, travel through each user interest-degree to every class service under every kind of context environmental;

B, judge whether the user has traveled through, if do not traveled through, execution in step c then; Otherwise, finish this ergodic process;

Whether c, judgement are finished in the traversal of certain class service at certain body and function family, then return step b if finish; Otherwise, execution in step d;

Whether d, judgement are finished at the context traversal of described individual consumer under certain class service, then return step c if finish; Otherwise, execution in step e;

E, judge that whether described individual consumer's interest-degree is zero, if, execution in step f then; Otherwise, execution in step j;

F, travel through this individual consumer place group every other user under this kind context environmental to the interest-degree of such service, and obtain this user place group identification, execution in step g then;

G, calculate described every other user effective mean value to the interest-degree of such service under this kind context environmental;

H, judge that whether described effective mean value be zero, if, execution in step i then; Otherwise, execution in step j;

I, individual consumer's preference value are the preference value of this user place group to such service, execution in step l then;

J, individual consumer's preference value are effective mean value; Execution in step l then;

K, individual consumer's preference value are this individual consumer's interest-degree, then execution in step l;

L, individual consumer's preference is stored in the corresponding database, return execution in step d then.