CN104537095A - Accurate information pushing method and system based on attraction model - Google Patents

Abstract

The invention discloses an accurate information pushing method based on an attraction model. The accurate information pushing method comprises the following steps that a system classifies recommended objects according to a preset rule; the interest level of a user in the recommended objects of the same category is calculated according to pre-defined interest parameters, the recommendable degree of the recommended objects of the same category is calculated according to pre-defined attention parameters, and the distance between the scenes of the recommended objects of the same category and the user is acquired; the attraction of the recommended objects of the same category to the user is calculated according to the interest level of the user in the recommended object of the same category, the recommendable degree of the recommended objects, the distance between the recommended objects and the user and the attraction model; the recommended objects are recommended to the user according to the sequence of the levels of attraction of the recommended objects to the user. The recommending mode is more accurate, it is guaranteed that the user acquires more accurate information conforming to needs, the attention of the user to unnecessary information is reduced, and it is guaranteed that the value of information acquired by the user is higher.

Description

A kind of accurate information method for pushing based on attractive force model and system

Technical field

The present invention relates to content and push field, refer more particularly to a kind of accurate information method for pushing based on attractive force model and system.

Background technology

The implementation of existing supplying system can be divided into two large classes usually: a kind of is the recommend method of commodity in e-commerce system, and another kind is the recommend method of good friend in social networks.Their common ground is based on the historical behavior data analysis of user in website, thus selects user's interested commodity of possibility or personage, recommends user.But there is some problem following in these method for pushing:

1, the algorithm realizing pushing is too simple, lacks the accurate evaluation to user preference.Such as many websites are, according to the browsing histories of user to the special article that website is shown or particular persons information, alternative product or people information are recommended user.In many cases, this kind of propelling movement mode is user's browsing history based on website accumulation or the data of operation history, the user that such as have purchased the first product also may buy the data of the second Related product, and the user having browsed A personage also may pay close attention to relevant B personage.

2, be generally only applied on Recommendations or personage, the scope of application is partially narrow.

3, many propelling movements are actually the Web broadcast model following rough type, and specific aim is not strong, easily causes spreading unchecked and wasting of information.Such as sharing and propagating of one section of article or a breaking news event, follow simple Web broadcast model exactly.Which results in spreading unchecked of information, have information greatly directly to be slatterned, because they do not cause interest and the concern of user.

4, be not suitable in the mobile Internet epoch to user's content recommendation.Instantly popular line lower pattern of reaching the standard grade needs commending system to have very strong Regional Analysis technology.Such as, a user moves in Beijing and Liang Ge city, Guangzhou, and suppertime needs to recommend the place of having a meal to him by mobile Internet.And existing method for pushing can not carry out analyzing and recommending according to the position at user place usually, lack Regional Analysis technology, be not thus suitable for this line and reach the standard grade lower pattern.

Therefore, prior art needs further to be improved.

Summary of the invention

Problem to be solved by this invention is, for the above-mentioned defect of prior art, a kind of accurate information method for pushing based on attractive force model and system are provided, solve the problem that existing method for pushing lacks accurate evaluation, narrow application range and shortage Regional Analysis technology to user preference.

The technical scheme that technical solution problem of the present invention adopts is as follows: a kind of accurate information method for pushing based on attractive force model, comprises the steps:

A, system are sorted out recommended object according to pre-defined rule;

B, calculate according to predefined interest parameter user to calculate recommended object in described class recommended degree to the interest level of the recommended object of a class and predefined concern parameter respectively, and obtain the distance in described class between recommended object scene and user;

C, utilize user to the recommended degree of recommended object in the interest level of the recommended object of described class, described class and and user between distance, to calculate in described class recommended object to the attractive force of user according to attractive force model;

D, according to the size order of the attractive force to user, object recommended in described class is recommended user.

Further, described step B specifically comprises:

B1, be some discrete time sections by system Time segments division, and recording user hour of log-on section and recommended object generation time section.

Further, described step B specifically also comprises:

B21, described interest parameter comprise interest decline factor-alpha, interest extender index β and user and pay close attention to frequency T to the recommended object of a class, and according to following formulae discovery user u to the interest level I of the recommended object of a class _{u, x}:

$I_{u,x}=C_1+\underset{i=0}{\overset{x-u}{\mathrm{\Σ}}}{\mathrm{\α}}^{x-u-i}{T_{u+i}}^{\mathrm{\β}}---\left(1\right)$

Wherein, user u is at t _utime period registers, C ₁for constant, T _u+ifor user u is at t _u+1time period pays close attention to frequency, 0 < α≤1, β>=1 to the recommended object of a class;

What B22, described concern parameter comprised the factor gamma that fails in time of recommended object, closeness extender index δ and recommended object is concerned frequency F, and according to recommended object in class described in following formulae discovery at t _xthe recommended degree R of time period _{a, x}:

$R_{a,x}=C_2+{\mathrm{\&Sigma;}}_{i=0}^{x-a}{\mathrm{\&gamma;}}^{x-a-i}{F_i}^{\mathrm{\&delta;}}---\left(2\right)$

Wherein, recommended object is at t _αtime period produces, C ₂for constant, F _ifor recommended object is concerned frequency, 0 < γ≤1, δ>=1 in each time period;

Described step C specifically also comprises:

Recommended object is calculated in described class to the attractive force of user according to following attractive force model:

$G_{\mathrm{aux}}=\frac{I_{u,x}\&times;R_{a,x}}{C_3+L^x}---\left(3\right)$

Wherein, G _auxfor recommended object is to the attractive force of user u, C ₃for constant, L is the distance in described class between recommended object scene and user, C ₃constant η is with η.

Further, when systems axiol-ogy pays close attention to new recommended object to user, system upgrades user to the interest level of this new recommended object generic by formula (1);

When systems axiol-ogy causes user to pay close attention to new recommended object, system upgrades the recommended degree of recommended object by formula (2).

Based on the accurate information supplying system of attractive force model, wherein, described system comprises:

Classifying module, for sorting out recommended object according to pre-defined rule;

First computing module, for calculating the recommended degree of recommended object in described class respectively to the interest level of the recommended object of a class and predefined concern parameter according to predefined interest parameter calculating user, and obtain the distance in described class between recommended object scene and user;

Second computing module, the user drawn for utilizing the first computing module to the recommended degree of recommended object in the interest level of the recommended object of described class, described class and and user between distance, to calculate in described class recommended object to the attractive force of user according to attractive force model;

Pushing module, recommends user for the size order according to the attractive force to user by object recommended in described class.

Further, described system also comprises memory module, for storing subscriber information, recommended object information, the interest level of user to all kinds of recommended object and the recommended degree of recommended object.

Further, described first computing module also comprises interest level computing module and can recommend degree computing module, wherein,

Interest level computing module, for calling algorithmic formula calculate user to the interest level of the recommended object of a class, wherein, described interest parameter comprises interest decline factor-alpha, interest extender index β and user and pays close attention to frequency T to the recommended object of a class, and user u is at t _utime period registers, C ₁for constant, T _u+ifor user u is at t _u+1time period pays close attention to frequency, 0 < α≤1, β>=1 to the recommended object of a class;

Degree computing module can be recommended, for calling algorithmic formula the recommended degree of recommended object in described class, wherein, what described concern parameter comprised the factor gamma that fails in time of recommended object, closeness extender index δ and recommended object is concerned frequency F, and recommended object is at t _αtime period produces, C ₂for constant, F _ifor recommended object is concerned frequency, 0 < γ≤1, δ>=1 in each time period;

Second computing module, also for calling algorithmic formula to calculate in described class recommended object to the attractive force of user, wherein, G _auxfor recommended object is to the attractive force of user u, C ₃for constant, L is the distance in described class between recommended object scene and user, C ₃constant η is with η.

Further, described system also comprises update module, for when systems axiol-ogy pays close attention to new recommended object to user, calls described interest level computing module and upgrades user to the interest level of this new recommended object generic; When systems axiol-ogy causes user to pay close attention to new recommended object, call the recommended degree that degree computing module can be recommended to upgrade recommended object.

Compared with prior art, accurate information method for pushing based on attractive force model provided by the invention and system, whether the interest and emotion model, recommended object and the distance of user and the focus degree of recommended object etc. that consider the mankind be attractive to user because usually considering an object, and determine whether recommend a certain object to user according to the size of this attractive force, more meet the model when mankind want obtaining information.Based on this computation model, the present invention, by extend to the information pushing of user than general electric business or the wider scope of social networks, except being applicable to traditional electrical business and social networks, is also applicable to the user information pushing in O2O epoch.This propelling movement mode can ensure that user obtains the information meeting demand more accurately, reduces user to the concern of unnecessary information, ensures that the value of the information that user obtains is larger.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the accurate information method for pushing based on attractive force model provided by the invention.

Fig. 2 is the recommended degree calculation flow chart of the accurate information method for pushing method based on attractive force model provided by the invention.

Fig. 3 is the coordinate schematic diagram of the system time discretize of the accurate information method for pushing based on attractive force model provided by the invention.

Fig. 4 is the example model figure of the accurate information method for pushing based on attractive force model provided by the invention.

Fig. 5 is the structured flowchart of the accurate information supplying system based on attractive force model provided by the invention.

Fig. 6 is the first computing module structured flowchart of the accurate information supplying system based on attractive force model provided by the invention.

Fig. 7 is each modular construction block diagram of the accurate information supplying system based on attractive force model provided by the invention.

Fig. 8 is calculating unit one structured flowchart of the accurate information supplying system based on attractive force model provided by the invention.

Fig. 9 is calculating unit two structured flowchart of the accurate information supplying system based on attractive force model provided by the invention.

Figure 10 is calculating unit three structured flowchart of the accurate information supplying system based on attractive force model provided by the invention.

Figure 11 is other auxiliary structured flowcharts of the accurate information supplying system based on attractive force model provided by the invention.

Embodiment

In order to make technical matters to be solved by this invention, technical scheme and beneficial effect clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

With reference to Fig. 1, a kind of accurate information method for pushing based on attractive force model provided by the invention, comprises the steps:

S100, system are sorted out recommended object according to pre-defined rule.

In method for pushing of the present invention, the scope of first clearly recommended object, likely can be become recommended object by the information that user pays close attention to, a such as breaking news event, the commodity that in e-commerce website or 020 website, businessman issues or service, article, picture or the multimedia messages of being shared or broadcasting is wanted for one section in social network sites, can a recommended user etc. in social network sites.Then sort out recommended object according to pre-defined rule, classifying method is unique, such as, can sort out by line with under line, sorts out by commodity and service, by the qualitative classification of Domestic News, classifies by the region of information and time.

S200, calculate according to predefined interest parameter user to calculate recommended object in described class recommended degree to the interest level of the recommended object of a class and predefined concern parameter respectively, and obtain the distance in described class between recommended object scene and user.

Principle of the present invention is when thinking that a recommended object occurs, it becomes certain polynomial expression inverse relation to the attractive force of user with the distance of user, become certain polynomial expression proportional relation with the class interest level of user to recommended object place, become certain polynomial expression proportional relation with recommended object recommended degree at that time.Therefore, by calculate user section sometime to the interest level of the recommended object of a class, user to described class in the recommended degree of recommended object and user and described class recommended object distance, then certain recommended object can be obtained sometime to the attractive force of user by the above-mentioned relation between three.

Step S200 specifically comprises the steps:

S210, be some discrete time sections by system Time segments division, and recording user hour of log-on section and recommended object generation time section.

The period discretize of system divided, be convenient to like this measure and record the time of user and the time of recommended object, the data gathered within the discrete time period are more suitable for computing machine and process.

As shown in Figure 2, step S200 further comprises the following step:

S221, described interest parameter comprise interest decline factor-alpha, interest extender index β and user and pay close attention to frequency T to the recommended object of a class, and according to following formulae discovery user u to the interest level I of the recommended object of a class _{u, x}:

$I_{u,x}=C_1+\underset{i=0}{\overset{x-u}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}^{x-u-i}{T_{u+i}}^{\mathrm{\&beta;}}---(1-1)$

Wherein, user u is at t _utime period registers, C ₁for constant, T _u+ifor user u is at t _u+1time period pays close attention to frequency, 0 < α≤1, β>=1 to the recommended object of a class;

As shown in Figure 3, in system time, suppose that user u is at t _utime period registers, then the time period of user is from t _uperiod starts timing, and a recommended object is at t _stime period produces, then the recommended object period is from t _speriod starts timing.If at discrete time section (t _u, t _u+1..., t _x) in, the frequency that user u pays close attention to the recommended object of a class is (T _u, T _u+1..., T _x), then user is at t _xthe interest level of period to a certain class event is:

$\left\{\begin{array}{cc}{I}_{u,x}=\mathrm{\&alpha;}{I}_{x-1}+{T_x}^{\mathrm{\&beta;}},& (x>u)\\ I_{u,x}={T_u}^{\mathrm{\&beta;}},& (x=u)\end{array}\right.---(1-2)$

Wherein α is the interest decline factor (0 < α≤1), and β is interest extender index (β >=1), can obtain after above formula distortion:

$I_{u,x}={\mathrm{\&Sigma;}}_{i=0}^{x-u}{\mathrm{\&alpha;}}^{x-u-i}{T_{u+i}}^{\mathrm{\&beta;}}---(1-3)$

Introduce user to the minimum value of certain object class interest level, with a constant C ₁represent, then can adopt following formula when system-computed:

$I_{u,x}=C_1+\underset{i=0}{\overset{x-u}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}^{x-u-i}{T_{u+i}}^{\mathrm{\&beta;}}---(1-1)$

What S222, described concern parameter comprised the factor gamma that fails in time of recommended object, closeness extender index δ and recommended object is concerned frequency F, and according to recommended object in class described in following formulae discovery at t _xthe recommended degree R of time period _{a, x}:

$R_{a,x}=C_2+{\mathrm{\&Sigma;}}_{i=0}^{x-a}{\mathrm{\&gamma;}}^{x-a-i}{F_i}^{\mathrm{\&delta;}}---(1-4)$

Wherein, recommended object is at t _αtime period produces, C ₂for constant, F _ifor recommended object is concerned frequency, 0 < γ≤1 in each time period, δ>=1;

Similar step S221, in system time, supposes that user u is at t _utime period registers, and a recommended object is at t _stime period produces, then the recommended object period is from t _speriod starts timing.If at t _xin time period, the frequency that the recommended object of class described in step S221 is concerned is F _x, then at t _xin period, the recommended degree of recommended object is:

$\left\{\begin{array}{cc}{R}_{a,x}=\mathrm{\&gamma;}{R}_{x-1}+{F_x}^{\mathrm{\&delta;}},& (x>\mathrm{\&alpha;})\\ R_{a,x}={F_a}^{\mathrm{\&beta;}},& (x=a)\end{array}\right.---(1-5)$

Wherein γ is the decline factor in time (γ) of recommended object, and δ is closeness extender index (δ >=1), can obtain after above formula distortion:

$R_{a,x}={\mathrm{\&Sigma;}}_{i=0}^{x-a}{\mathrm{\&gamma;}}^{x-a-i}{F_i}^{\mathrm{\&delta;}}---(1-6)$

The average valuation of certain period object class that drawing-in system has calculated, uses C ₂represent, then can adopt following formula when system-computed:

$R_{a,x}=C_2+{\mathrm{\&Sigma;}}_{i=0}^{x-a}{\mathrm{\&gamma;}}^{x-a-i}{F_i}^{\mathrm{\&delta;}}---(1-4)$

S300, utilize user to the recommended degree of recommended object in the interest level of the recommended object of described class, described class and and user between distance, to calculate in described class recommended object to the attractive force of user according to attractive force model;

Described step S300 specifically also comprises:

Recommended object is calculated in described class to the attractive force of user according to following attractive force model:

$G_{\mathrm{aux}}=\frac{I_{u,x}\&times;R_{a,x}}{C_3+L^x}---(1-7)$

Wherein, G _auxfor recommended object is to the attractive force of user u, C ₃for constant, L is the distance in described class between recommended object scene and user, C ₃constant η is with η.

System obtains t _xthe distance L of time period user u and recommended object a, and the user that integrating step S200 calculates is at t _xperiod is to the interest level I of a certain class event _{u, x}, and at t _xthe recommended degree R of period recommended object _{a, x}, according to attractive force model to calculate in described class recommended object to the attractive force of user, wherein C ₃be a certain normal number of default, avoid occurring except 0 mistake.

Distance in above-mentioned computing formula can also be made adjustment according to actual conditions, such as, if user moves B ground from A at A after the long period of having lived, may still to A ground commodity, service, Domestic News or personage interested, so at this moment just can not be calculated by the actual range on user and A ground merely, size by regulating parameter η changes the size of denominator in formula (1-7), thus the attractive force size of the recommended object calculated to user more tallies with the actual situation.

S400, according to the size order of the attractive force to user, object recommended in described class is recommended user.

The attractive force of each recommended object to user calculated by step S300 sorts from big to small, and what attractive force was large is preferentially pushed to user.

In above-mentioned implementation process, when systems axiol-ogy pays close attention to new recommended object to user, system upgrades the interest level of user to this new recommended object generic by formula (1-1); When systems axiol-ogy causes user to pay close attention to new recommended object, system upgrades the recommended degree of recommended object by formula (1-4).Perform step S300 and step S400 again after having upgraded, calculate the recommended object after renewal to the attractive force of user, and be pushed to user by order from big to small.

With an example, said process is described below, as shown in Figure 4, recommended object A1-A9 is divided into C1, C2 and C3 tri-class, if user 1 is at t _utime period registers, and recommended object A4 is at t _stime period produces.Now to calculate recommended object A4 to the attractive force of user 1, then calculate the interest level I of user's 1 pair of C1 class according to formula (1-1) _{c1_n}, and calculate the recommended degree R of A4 within certain time period by formula (1-4) _{a4_n}, then combine the place of A4 generation and the distance L of user 1 that calculate, utilize formula (1-7) to obtain the attractive force size of A4 to user 1.

Similarly, other recommended objects can be calculated in C1 class according to the method described above to the attractive force size of user 1, and recommended object, to the attractive force size of user 1, sorts to each recommended object A1-A9, is pushed to user 1 by priority ranking from big to small in C2 class and C3 class.

As Fig. 5, based on above-mentioned accurate method for pushing, present invention also offers a kind of accurate information supplying system based on attractive force model, comprising:

Classifying module 10, for sorting out recommended object according to pre-defined rule;

First computing module 20, for calculating the recommended degree of recommended object in described class respectively to the interest level of the recommended object of a class and predefined concern parameter according to predefined interest parameter calculating user, and obtain the distance in described class between recommended object scene and user;

Second computing module 30, the user drawn for utilizing the first computing module 20 to the recommended degree of recommended object in the interest level of the recommended object of described class, described class and and user between distance, to calculate in described class recommended object to the attractive force of user according to attractive force model;

Pushing module 40, recommends user for the size order according to the attractive force to user by object recommended in described class.

Described system also comprises memory module 50, for storing subscriber information, recommended object information, the interest level of user to all kinds of recommended object and the recommended degree of recommended object.

As shown in Figure 6, described first computing module 20 also comprises:

Interest level computing module 21, for calling algorithmic formula calculate user to the interest level of the recommended object of a class, wherein, described interest parameter comprises interest decline factor-alpha, interest extender index β and user and pays close attention to frequency T to the recommended object of a class, and user u is at t _utime period registers, C ₁for constant, T _u+ifor user u is at t _u+1time period pays close attention to frequency, 0 < α≤1, β>=1 at the recommended object of a class;

Degree computing module 22 can be recommended, for calling algorithmic formula the recommended degree of recommended object in described class, wherein, what described concern parameter comprised the factor gamma that fails in time of recommended object, closeness extender index δ and recommended object is concerned frequency F, and recommended object is at t _αtime period produces, C ₂for constant, F _ifor recommended object is concerned frequency, 0 < γ≤1, δ>=1 in each time period;

Second computing module 30, also for calling algorithmic formula to calculate in described class recommended object to the attractive force of user, wherein, G _auxfor recommended object is to the attractive force of user u, C ₃for constant, L is the distance in described class between recommended object scene and user, C ₃constant η is with η.

Described system also comprises update module 60, for when systems axiol-ogy pays close attention to new recommended object to user, calls described interest level computing module 21 and upgrades the interest level of user to this new recommended object generic; When systems axiol-ogy causes user to pay close attention to new recommended object, call the recommended degree that degree computing module 22 can be recommended to upgrade recommended object.

As shown in Figure 7, in computer system implementation, above-mentioned functions module is equivalent to different calculating units or equipment, specifically comprises calculating unit 1, calculating unit 22, calculating unit 33, database or memory device 5 and other auxiliaries 4.As shown in Figure 8, calculating unit 1 is for upgrading the interest level of user 6 to this new recommended object 7 generic, when systems axiol-ogy pays close attention to new recommended object 7 to user 6, update module calls calculating unit 1, read from database or memory device 5 or from the storage area of self program, obtain the interest level data of passing user 6 to recommended object 7 place class, then the interest level of user 6 to recommended object 7 class is recalculated according to formula (1-1), and new user 6 interest level data are write back in the storage area of database or memory device 5 or self program.

As shown in Figure 9, calculating unit 22 is for upgrading the recommended degree of recommended object, when systems axiol-ogy causes user 6 to pay close attention to new recommended object 7, call calculating unit 22, read from database or memory device 5 or obtain the data relevant to degree can be recommended from the storage area of self program, recalculate the recommended degree of recommended object 7 again according to formula (1-4), and the related data of new user 6 and recommended object 7 is write back in the storage area of database or memory device 5 or self program.

As shown in Figure 10, calculating unit 33 mainly performs the computing of formula (1-7), when user 6 accessing system or system produce a recommended object 7 for a user 6 or a collection of user, call calculating unit 33, read or read from the storage area of self program from database, memory device 5 result that a series of or whole calculating unit 1 and calculating unit 22 draw, read relevant recommendation record (i.e. the relation of recommended object and user) simultaneously, and read distance or the geography information of user 6 and recommended object 7.For each user, filtration needs according to system filters out the object of recommended mistake, calculate user or batch user separately and the distance Lu of each recommended object, a, then utilize formula (1-7) for each chooses user and corresponding a series of recommended object 7 to calculate the attractive force of recommended object 7 couples of users, and above-mentioned result of calculation is sorted from big to small, call network network equipment simultaneously, by above-mentioned order, result is pushed to user, finally propelling movement record (i.e. the relation of recommended object and user) is saved on memory device 5.

Memory device is for depositing the intermediate data and operation result that produce in calculating, other auxiliaries are used for some subsidiary functions (as Figure 11), comprise input and output, call or trigger each calculating unit, preserve result of calculation and intermediate data, with the function such as other of user 6 are mutual.Part as received outside input or system exports and inputs as program, call when user 6 pays close attention to recommended object 7 or trigger calculating unit 1, recommended object 7 causes and to call during new concern or to trigger calculating unit 22, requirement is recommended when user 6 accessing system initiatively causes, or when system produces and requires for the recommendation of user 6, call calculating unit 33, when receiving the geographical location information of user 6 transmission, or when new recommended object 7 produces, the geographical location information that user 6 geographical location information or recommended object 7 produce is saved in memory device 5, also for realizing other interactive functions with user.

Separate between each calculating unit above-mentioned, both can by a program intrinsic call in a system, also multiple parts and process can run when system is cluster simultaneously, and the data that all parts calculates can be kept in memory device, which enhance the efficiency of data processing, decrease the time of system wait, arithmetic speed is accelerated, thus faster for user pushes various interested object information.

The accurate information method for pushing based on attractive force model that theres is provided of the present invention and system, adopt account form more accurately to push, and more meets the model when mankind want obtaining information.The present invention, by extend to the information pushing of user than general electric business or the wider scope of social networks, except being applicable to traditional electrical business and social networks, is also applicable to the user information pushing in O2O epoch.Ensure that user obtains the information meeting demand more accurately, reduce user to the concern of unnecessary information.Ensure that the value of the information that user obtains is larger.

Should be understood that, application of the present invention is not limited to above-mentioned citing, for those of ordinary skills, can be improved according to the above description or convert, and all these improve and convert the protection domain that all should belong to claims of the present invention.

Claims (8)

1., based on an accurate information method for pushing for attractive force model, it is characterized in that, comprise the steps:

A, system are sorted out recommended object according to pre-defined rule;

B, calculate according to predefined interest parameter user to calculate recommended object in described class recommended degree to the interest level of the recommended object of a class and predefined concern parameter respectively, and obtain the distance in described class between recommended object scene and user;

C, utilize user to the recommended degree of recommended object in the interest level of the recommended object of described class, described class and and user between distance, to calculate in described class recommended object to the attractive force of user according to attractive force model;

D, according to the size order of the attractive force to user, object recommended in described class is recommended user.

2. the accurate information method for pushing based on attractive force model according to claim 1, it is characterized in that, described step B specifically comprises:

B1, be some discrete time sections by system Time segments division, and recording user hour of log-on section and recommended object generation time section.

3. the accurate information method for pushing based on attractive force model according to claim 2, it is characterized in that, described step B specifically also comprises:

B21, described interest parameter comprise interest decline factor-alpha, interest extender index β and user and pay close attention to frequency T to the recommended object of a class, and according to following formulae discovery user u to the interest level I of the recommended object of a class _{u, x}:

Wherein, user u is at t _utime period registers, C ₁for constant, T _u+ifor user u is at t _u+1time period pays close attention to frequency, 0 < α≤1, β>=1 to the recommended object of a class;

What B22, described concern parameter comprised the factor gamma that fails in time of recommended object, closeness extender index δ and recommended object is concerned frequency F, and according to recommended object in class described in following formulae discovery at t _xthe recommended degree R of time period _{a, x}:

Wherein, recommended object is at t _atime period produces, C ₂for constant, F _ifor recommended object is concerned frequency, 0 < γ≤1, δ>=1 in each time period;

Described step C specifically also comprises:

Recommended object is calculated in described class to the attractive force of user according to following attractive force model:

Wherein, G _auxfor recommended object is to the attractive force of user u, C ₃for constant, L is the distance in described class between recommended object scene and user, C ₃constant η is with η.

4. the accurate information method for pushing based on attractive force model according to claim 3, is characterized in that,

When systems axiol-ogy pays close attention to new recommended object to user, system upgrades user to the interest level of this new recommended object generic by formula (1);

When systems axiol-ogy causes user to pay close attention to new recommended object, system upgrades the recommended degree of recommended object by formula (2).

5. based on an accurate information supplying system for attractive force model, it is characterized in that, described system comprises:

Classifying module, for sorting out recommended object according to pre-defined rule;

First computing module, for calculating the recommended degree of recommended object in described class respectively to the interest level of the recommended object of a class and predefined concern parameter according to predefined interest parameter calculating user, and obtain the distance in described class between recommended object scene and user;

Second computing module, the user drawn for utilizing the first computing module to the recommended degree of recommended object in the interest level of the recommended object of described class, described class and and user between distance, to calculate in described class recommended object to the attractive force of user according to attractive force model;

Pushing module, recommends user for the size order according to the attractive force to user by object recommended in described class.

6. the accurate information supplying system based on attractive force model according to claim 5, it is characterized in that, described system also comprises memory module, for storing subscriber information, recommended object information, the interest level of user to all kinds of recommended object and the recommended degree of recommended object.

7. the accurate information supplying system based on attractive force model according to claim 5, is characterized in that, described first computing module also comprises interest level computing module and can recommend degree computing module, wherein,

Interest level computing module, for calling algorithmic formula calculate user to the interest level of the recommended object of a class, wherein, described interest parameter comprises interest decline factor-alpha, interest extender index β and user and pays close attention to frequency T to the recommended object of a class, and user u is at t _utime period registers, C ₁for constant, T _u+ifor user u is at t _u+1time period pays close attention to frequency, 0 < α≤1, β>=1 to the recommended object of a class;

Degree computing module can be recommended, for calling algorithmic formula the recommended degree of recommended object in described class, wherein, what described concern parameter comprised the factor gamma that fails in time of recommended object, closeness extender index δ and recommended object is concerned frequency F, and recommended object is at t _atime period produces, C ₂for constant, F _ifor recommended object is concerned frequency, 0 < γ≤1, δ>=1 in each time period;

Second computing module, also for calling algorithmic formula to calculate in described class recommended object to the attractive force of user, wherein, G _auxfor recommended object is to the attractive force of user u, C ₂for constant, L is the distance in described class between recommended object scene and user, C ₂constant η is with η.

8. the accurate information supplying system based on attractive force model according to claim 7, it is characterized in that, described system also comprises update module, for when systems axiol-ogy pays close attention to new recommended object to user, call described interest level computing module and upgrade user to the interest level of this new recommended object generic; When systems axiol-ogy causes user to pay close attention to new recommended object, call the recommended degree that degree computing module can be recommended to upgrade recommended object.