CN106228403A - A kind of user based on step analysis algorithm is worth methods of marking and system - Google Patents

Abstract

The invention discloses a kind of user based on step analysis algorithm and be worth methods of marking and system, relate to the user in net cast and be worth scoring field.The method includes: according to the behavioural information of user, build the analytic hierarchy structure that user is worth；Carry out step analysis to every layer, calculate each first class index to the weighing factor of user's comprehensive value, calculate each two-level index weighing factor to corresponding first class index；Each two-level index weight to user's comprehensive value is calculated according to calculated weighing factor；The mode using weighted sum again calculates the comprehensive value score of each user；Finally according to the comprehensive value score of each user calculated, choose corresponding score interval, mark off the value grade of each user according to different score intervals.The present invention can realize the automatic scoring of user's comprehensive value, and scoring process is accurate, efficient, save human cost.

Description

A kind of user based on step analysis algorithm is worth methods of marking and system

Technical field

The present invention relates to the user in net cast and be worth scoring field, be specifically a kind of based on step analysis algorithm User be worth methods of marking and system.

Background technology

Along with developing rapidly of live industry, enriching of platform service, userbase presents the trend of being skyrocketed through.User exists Produce different various actions on platform, bring different value to website, such as gifts, charging information etc., how to evaluate The comprehensive value score of user, which is high-value user, and which is low value user, facilitates marketing personnel according to user not With being worth grade, plan different marketing program with maintaining and keep strategy, be a problem to be solved.

At present, each big website when carrying out comprehensive value scoring to current user, or utilization is manually rule of thumb advised Then carry out screening high value, low value user, not yet have suitable algorithm to carry out the evaluation of user's comprehensive value.Further, in reality In the operation of border, entirely by manually using manual screening high value, the method for low value user, often with subjective largely Property so that evaluation criterion disunity；It addition, under the scene of mass data, user behavior data often dimension is many, data volume Greatly, manually pass judgment on the user gradation criteria for classifying is the most inaccurate, coverage rate is the most not high enough, repetitive work is also easily caused Error, and handling time is longer, divides inefficient, and human cost is bigger.

Summary of the invention

The invention aims to overcome the deficiency of above-mentioned background technology, it is provided that a kind of use based on step analysis algorithm Family is worth methods of marking and system, can be based on the user's value level analytical structure built, according to the behavior quantitative Analysis of user Go out the comprehensive value score of user, and scoring process is accurate, efficient, save human cost.

For reaching object above, the present invention provides a kind of user based on step analysis algorithm to be worth methods of marking, including Following steps:

Step S1: according to the behavioural information of user, build the analytic hierarchy structure that user is worth, the level that this user is worth Analytical structure is three-decker: ground floor is user's comprehensive value, and the second layer is that multiple one-levels corresponding to user's comprehensive value refer to Mark, third layer is multiple two-level index that each first class index is corresponding, proceeds to step S2；

Step S2: the second layer is carried out step analysis, calculates each first class index weighing factor to user's comprehensive value； Third layer is carried out step analysis, calculates each two-level index weighing factor to corresponding first class index, proceed to step S3；

Step S3: according to each first class index to the weighing factor of user's comprehensive value and each two-level index to accordingly The weighing factor of first class index, calculates each two-level index weight W to user's comprehensive value ', define W_e' represent e two The level index weight to user's comprehensive value, e=1,2 ..., k, k is the sum of two-level index, proceeds to step S4；

Step S4: according to each two-level index calculated weight W to user's comprehensive value ', use the side of weighted sum Formula calculates comprehensive value score S of each user, defines S_zRepresent the comprehensive value score of z position user, z=1,2 ..., p, p It is represented as maximum number of user amount, proceeds to step S5；

Step S5: according to comprehensive value score S of each user calculated, chooses corresponding score interval；According to difference Score interval mark off the value grade of each user.

On the basis of technique scheme, described in step S1, multiple first class index include: liveness, the consuming capacity, Supplement ability with money；The plurality of two-level index includes: the user corresponding with liveness watches duration, user watches natural law, Yong Huguan Seeing that room number, user send barrage number, the user corresponding with the consuming capacity gives virtual present amount, user gives virtual present room Between number, the user recharge amount corresponding with the ability of supplementing with money, average recharge amount every time, supplement natural law with money.

On the basis of technique scheme, in step S2, calculate the impact on user's comprehensive value of each first class index Weight, calculate each two-level index the process of the weighing factor of corresponding first class index all included following operation:

Step 1, selective goal: select the index needing to calculate weighing factor in this step analysis；

Step 2, development of judgment matrix: by the index selected is compared, two-by-two according to 1-9 scaling law deliberated index Between relative importance grade, construct the judgment matrix A of selected index:

$A=\left[\begin{array}{cccc}1& a_{12}& ...& a_{1n}\\ a_{21}& 1& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& 1\end{array}\right]$

A in definition judgment matrix A_ijRepresent the importance comparative result of index i and index j, and a_ij=1/a_ji, i=1, 2 ..., n, j=1,2 ..., n, n represent the number of selected index, the i.e. exponent number of judgment matrix A；

Step 3, calculating weighing factor: current judgment matrix A is pressed row normalization, obtains normalization matrix B, matrix B Inb_ijFor the i-th row normalized value of j row element in matrix A；Normalization matrix B is obtained c by row summation_i,By c_iIt is normalized, obtains the weighing factor w that selected index is corresponding_i, w_iI-th index selected by expression is corresponding Weighing factor, its computing formula is:

Step 4, calculating characteristic vector and eigenvalue of maximum thereof: according to the weighing factor w of currently available selected index_i, Obtain the characteristic vector W of current judgment matrix A, W=(w₁, w₂..., w_n)^T, and obtain the eigenvalue of maximum that characteristic vector W is corresponding λ_max,

Step 5, check and adjust judgment matrix, drawing final weighing factor: according to eigenvalue of maximum λ_max, calculate Coincident indicator CI of current judgment matrix A, its computing formula is:Calculate according to coincident indicator CI The Consistency Ratio CR of current judgment matrix A, its computing formula is: CR=CI/RI, RI are corresponding average of judgment matrix exponent number Coincident indicator data；The Consistency Ratio CR that inspection calculates, whether less than specifying threshold value, if so, shows judgment matrix A's Discordance degree in permissible range, then directly the most corresponding by obtaining selected index in step 3 weighing factor w_iAs selected The final weighing factor that index is corresponding；Otherwise, after adjusting judgment matrix A, repeated execution of steps 3 to step 5, until judgment matrix The CR of A is less than specifying threshold value, then using weighing factor corresponding for the selected index of latest computed as corresponding final of selected index Weighing factor.

On the basis of technique scheme, in step S3, W_e' computing formula be: Wherein w_f Representing the f first class index weighing factor to user's comprehensive value, f=1,2 ..., m, m is the sum of first class index, v_efTable Show the e two-level index weighing factor to f first class index.

On the basis of technique scheme, in step S4, S_zComputing formula be: S_z=W₁’*X_z1’+W₂’*X_z2’+…+ W_e’*X_ze', X_ze' represent value X of e the two-level index of z position user_zeValue after normalized process, X_ze' calculating public Formula is:

${X_{ze}}^{,}=\frac{X_{ze}-min(X_{1e},X_{2e},...,X_{pe})}{max(X_{1e},X_{2e},...,X_{pe})-\min (X_{1e},X_{2e},...,X_{pe})};$

Max (X in above-mentioned formula_1e, X_2e..., X_pe) represent the maximum of e two-level index in all users；min (X_1e, X_2e..., X_pe) represent and choose the minima of e two-level index in all users.

The present invention also provides for a kind of user based on step analysis algorithm simultaneously and is worth marking system, ties including step analysis Structure builds module, Index Influence weight computation module, two-level index comprehensive value weight computation module, the calculating of user's comprehensive value Module, value grade classification module；

Described analytic hierarchy structure builds module and is used for: according to the behavioural information of user, the level building user's value divides Analysis structure, the analytic hierarchy structure that this user is worth is three-decker: ground floor is user's comprehensive value, and the second layer is that user combines Closing and be worth corresponding multiple first class index, third layer is multiple two-level index that each first class index is corresponding；

Described Index Influence weight computation module is used for: the second layer in user's value level analytical structure is carried out level Analyze, calculate each first class index weighing factor to user's comprehensive value；To the 3rd in user's value level analytical structure Layer carries out step analysis, calculates each two-level index weighing factor to corresponding first class index；

Described two-level index comprehensive value weight computation module is used for: according to each first class index to user's comprehensive value Weighing factor and each two-level index weighing factor to corresponding first class index, calculate each two-level index valency comprehensive to user Weight W of value ', define W_e' representing the e two-level index weight to user's comprehensive value, e=1,2 ..., k, k is two grades and refers to Target sum；

Described user's comprehensive value computing module is used for: calculate according to two-level index comprehensive value weight computation module Each two-level index weight W to user's comprehensive value ', use the mode of weighted sum to calculate the comprehensive value score of each user S, defines S_zRepresenting the comprehensive value score of z position user, z=1,2 ..., p, p is represented as maximum number of user amount；

Described value grade classification module is used for: each user's calculated according to user's comprehensive value computing module is comprehensive It is worth score S, chooses corresponding score interval；The value grade of each user is marked off according to different score intervals.

On the basis of technique scheme, the plurality of first class index includes: liveness, the consuming capacity, supplement ability with money； The plurality of two-level index includes: the user corresponding with liveness watches duration, user watches natural law, user watches room number, User sends barrage number, and the user corresponding with the consuming capacity gives virtual present amount, user gives virtual present room number, and fills User's recharge amount corresponding to value ability, average recharge amount every time, supplement natural law with money.

On the basis of technique scheme, described Index Influence weight computation module calculates each first class index to user The weighing factor of comprehensive value, calculate each two-level index the process of the weighing factor of corresponding first class index is all included following behaviour Make:

Selective goal: select the index needing to calculate weighing factor in this step analysis；

Development of judgment matrix: by the index selected is compared, two-by-two between 1-9 scaling law deliberated index Relative importance grade, constructs the judgment matrix A of selected index:

$A=\left[\begin{array}{cccc}1& a_{12}& ...& a_{1n}\\ a_{21}& 1& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& 1\end{array}\right]$

A in definition judgment matrix A_ijRepresent the importance comparative result of index i and index j, and a_ij=1/a_ji, i=1, 2 ..., n, j=1,2 ..., n, n represent the number of selected index, the i.e. exponent number of judgment matrix A；

Calculate weighing factor: current judgment matrix A is pressed row normalization, obtains normalization matrix B, in matrix Bb_ijFor the i-th row normalized value of j row element in matrix A；Normalization matrix B is obtained c by row summation_i,By c_iIt is normalized, obtains the weighing factor w that selected index is corresponding_i, w_iI-th index selected by expression is corresponding Weighing factor, its computing formula is:

Calculate characteristic vector and eigenvalue of maximum thereof: according to the weighing factor w of currently available selected index_i, worked as The characteristic vector W of front judgment matrix A, W=(w₁, w₂..., w_n)^T, and obtain the eigenvalue of maximum λ that characteristic vector W is corresponding_max,

Check and adjust judgment matrix, drawing final weighing factor: according to eigenvalue of maximum λ_max, calculate and currently sentence Coincident indicator CI of disconnected matrix A, its computing formula is:Calculate according to coincident indicator CI and currently sentence The Consistency Ratio CR of disconnected matrix A, its computing formula is: CR=CI/RI, RI are the average homogeneity that judgment matrix exponent number is corresponding Achievement data；The Consistency Ratio CR that inspection calculates, whether less than specifying threshold value, if so, shows that judgment matrix A's is inconsistent Property degree in permissible range, then directly by weighing factor w corresponding for currently available selected index_iCorresponding as selected index Final weighing factor；Otherwise, after adjusting judgment matrix A, repeated execution of steps 3 to step 5, until the CR of judgment matrix A is less than Specify threshold value, then using weighing factor corresponding for the selected index of latest computed as final weighing factor corresponding to selected index.

On the basis of technique scheme, described two-level index comprehensive value weight computation module calculates W_e' formula For:Wherein w_fRepresent the f first class index weighing factor to user's comprehensive value, f=1,2 ..., M, m are the sum of first class index, v_efRepresent the e two-level index weighing factor to f first class index.

On the basis of technique scheme, described user's comprehensive value computing module calculates S_zFormula be: S_z=W₁’* X_z1’+W₂’*X_z2’+…+W_e’*X_ze', X_ze' represent value X of e the two-level index of z position user_zeAfter normalized process Value, X_ze' computing formula be:

${X_{ze}}^{,}=\frac{X_{ze}-min(X_{1e},X_{2e},...,X_{pe})}{max(X_{1e},X_{2e},...,X_{pe})-\min (X_{1e},X_{2e},...,X_{pe})};$

Max (X in above-mentioned formula_1e, X_2e..., X_pe) represent the maximum of e two-level index in all users；min (X_1e, X_2e..., X_pe) represent and choose the minima of e two-level index in all users.

The beneficial effects of the present invention is:

(1) present invention is when carrying out user and being worth scoring, first according to the behavioural information of user, builds the level that user is worth Analytical structure；Then step analysis is carried out, the weighing factor of each index in calculating every layer；Calculate then according to weighing factor All two-level index weight to user's comprehensive value；The mode using weighted sum again calculates the comprehensive value of each user and obtains Point；Comprehensive value score finally according to user marks off the value grade of each user.

Compared with prior art, the present invention can behavior based on user, the comprehensive value score of quantitative Analysis user；Then Comprehensive value score according to user, marks off different high, normal, basic value grades；Finally make the marketing personnel can be according to different User is worth grade and designs different marketing and retention tactics, it is simple to orientation is marketed and maintains.Whole operation not only makes to use The value scoring process at family becomes more intelligent, automatization；And user is worth that the quality of scoring is high, efficiency high, reliable Property strong, effectively save human cost, Consumer's Experience is effective.

(2), during the weighing factor of the present invention each index in every time calculating every layer, the judgment matrix of structure is used.Every time After utilizing judgment matrix to calculate the weighing factor of corresponding index, judgment matrix can be carried out repeated examinations and adjustment, until sentencing The discordance degree (i.e. CR ＜ 0.1) in permissible range of disconnected matrix, just draws final weighing factor, thus is effectively improved User is worth the accuracy of scoring.

(3) present invention is when calculating comprehensive value score S of each user, uses after normalized process two grades to refer to Target value X_ze' rather than directly use two-level index value X_ze, thus the dimension being prevented effectively from the value because of two-level index differs Cause and result of calculation is impacted, and then ensure that the reliability of calculating.

Accompanying drawing explanation

Fig. 1 is the flow chart that in the embodiment of the present invention, user based on step analysis algorithm is worth methods of marking；

Fig. 2 is the schematic diagram of the analytic hierarchy structure that user is worth in the embodiment of the present invention；

Fig. 3 is the flow chart of the weighing factor calculating monolayer index in the embodiment of the present invention；

Fig. 4 is the structured flowchart that in the embodiment of the present invention, user based on step analysis algorithm is worth marking system.

Detailed description of the invention

Below in conjunction with the accompanying drawings and specific embodiment the present invention is described in further detail.

Shown in Figure 1, the embodiment of the present invention provides a kind of user based on step analysis algorithm to be worth methods of marking, bag Include following steps:

Step S1: according to the behavioural information of user, build the analytic hierarchy structure that user is worth, the level that this user is worth Analytical structure is three-decker: ground floor is user's comprehensive value, and the second layer is that multiple one-levels corresponding to user's comprehensive value refer to Mark, third layer is multiple two-level index that each first class index is corresponding, proceeds to step S2.

Shown in Figure 2, in the present embodiment, the plurality of first class index includes: liveness, the consuming capacity, supplement ability with money； The plurality of two-level index includes: the user corresponding with liveness watches duration, user watches natural law, user watches room number, User sends barrage number, and the user corresponding with the consuming capacity gives virtual present amount, user gives virtual present room number, and fills User's recharge amount corresponding to value ability, average recharge amount every time, supplement natural law with money.I.e. as in figure 2 it is shown, user be worth level Analytical structure is: ground floor is user's comprehensive value；The second layer respectively liveness, the consuming capacity, supplement ability with money；Third layer is divided User that Wei not be corresponding with liveness watches duration, user watches natural law, user watches room number, user sends barrage number, with The user that the consuming capacity are corresponding gives virtual present amount, user gives virtual present room number, the user corresponding with the ability of supplementing with money Recharge amount, average recharge amount every time, supplement natural law with money.

Step S2: the second layer is carried out step analysis, calculates each first class index weighing factor to user's comprehensive value； Third layer is carried out step analysis, calculates each two-level index weighing factor to corresponding first class index, proceed to step S3.

It is understood that the analytic hierarchy structure that the user owing to building in the present embodiment is worth is as in figure 2 it is shown, have 3 first class index, each first class index, the most respectively to there being multiple two-level index, therefore, when carrying out step S2, need to be carried out altogether 4 analytical calculations, just can calculate the weighing factor of 12 indexs (3 first class index, 9 two-level index), and concrete operations are such as Under:

Step S201, for the first time: the second layer carries out step analysis, calculates liveness and weighs the impact of user's comprehensive value Weight, the consuming capacity to the weighing factor of user's comprehensive value, supplement the ability weighing factor to user's comprehensive value with money；

Step S202, for the second time: third layer carries out step analysis, calculates user and watches duration and weigh the impact of liveness Weight, user watch natural law and watch weighing factor, the user of liveness room number and weighing factor, the user of liveness are sent bullet The several weighing factors to liveness of curtain；

Step S203, for the third time: third layer carries out step analysis, calculates user and gives virtual present amount to the consuming capacity Weighing factor, user give the virtual present room number weighing factor to the consuming capacity；

Step S204, the 4th time: third layer is carried out step analysis, calculates the impact on the ability of supplementing with money of user's recharge amount Weight, average recharge amount every time to the weighing factor of the ability of supplementing with money, supplement the natural law weighing factor to the ability of supplementing with money with money.

Specifically, shown in Figure 3, carry out step analysis every time, when calculating the weighing factor of corresponding index, it calculates Method is similar to, and specifically includes following operation:

Step 1, selective goal: select the index needing to calculate weighing factor in this step analysis (as carried out for the first time During step analysis, selected index is liveness, the consuming capacity, supplements ability with money).

Step 2, development of judgment matrix: by the index selected is compared, two-by-two according to 1-9 scaling law deliberated index Between relative importance grade, construct the judgment matrix A of selected index:

$A=\left[\begin{array}{cccc}1& a_{12}& ...& a_{1n}\\ a_{21}& 1& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& 1\end{array}\right]$

A in definition judgment matrix A_ijRepresent the importance comparative result of index i and index j, and a_ij=1/a_ji, i=1, 2 ..., n, j=1,2 ..., n, n be represented as the number of selected index, the i.e. exponent number of judgment matrix A；Wherein, a_ijValue be with In lower 91: 1/9,1/7,1/5,1/3,1/1,3/1,5/1,7/1,9/1, numerical value is the biggest, and to represent significance level the highest, example Such as a_ij=1/9, represent that index i is the most inessential to index j, a_ij=9/1, represent that index i is extremely important to index j.a_ijTool Body value given a mark by multiple business experts after synthetic determination.

Step 3, calculating weighing factor: by current judgment matrix A by row normalization (i.e. column element sum is 1), obtain Normalization matrix B, in matrix Bb_ijFor the i-th row normalized value of j row element in matrix A；By normalization matrix B C is obtained by row summation_i,By c_iIt is normalized, obtains the weighing factor w that selected index is corresponding_i(that is, selected The weighing factor that i-th index is corresponding), its computing formula is:

Step 4, calculating characteristic vector and eigenvalue of maximum thereof: according to the weighing factor w of currently available selected index_i, Obtain the characteristic vector W of current judgment matrix A, W=(w₁, w₂..., w_n)^T, and obtain the eigenvalue of maximum that characteristic vector W is corresponding λ_max,

Step 5, check and adjust judgment matrix, drawing final weighing factor: according to eigenvalue of maximum λ_max, calculate Coincident indicator CI of current judgment matrix A, its computing formula is:(CI value is the biggest, shows that judgment matrix A is inclined The biggest from the degree of crash consistency；CI value is the least, shows that the concordance of judgment matrix A is the best)；Count according to coincident indicator CI Calculating the Consistency Ratio CR drawing current judgment matrix A, its computing formula is: CR=CI/RI, RI are that judgment matrix exponent number is corresponding Average homogeneity achievement data, this average homogeneity achievement data concrete value ginseng be shown in Table 1:

The concrete data that table 1, average homogeneity index (RI) are corresponding

n	1	2	3	4	5	6	7	8	9	10	11
												RI	0	0	0.58	0.9	1.12	1.24	1.32	1.41	1.45	1.49	1.51

Such as, if judgment matrix exponent number is 3, then value 0.58 corresponding during exponent number n=3 during RI chooses table 1；

Whether the Consistency Ratio CR that inspection calculates is less than is specified threshold value (specifying threshold value in the present embodiment is 0.1), if Be, show the discordance degree of judgment matrix A in permissible range, then directly the most corresponding by step 3 obtains selected index Weighing factor w_iAs final weighing factor corresponding to selected index (will the characteristic vector W of current judgment matrix A as weight Vector)；Otherwise, after adjusting judgment matrix A, repeated execution of steps 3 to step 5, until the discordance degree of judgment matrix A exists In permissible range (i.e. CR is less than specifying threshold value 0.1), then using weighing factor corresponding for the selected index of latest computed as selected The final weighing factor that index is corresponding.

Step S3: according to each first class index to the weighing factor of user's comprehensive value and each two-level index to accordingly The weighing factor of first class index, calculates each two-level index weight W to user's comprehensive value ', define W_e' represent e two The level index weight to user's comprehensive value, e=1,2 ..., k, k is the sum of two-level index, proceeds to step S4.

Wherein, W_e' computing formula be:w_fRepresent that the f first class index is to user's comprehensive value Weighing factor, f=1,2 ..., m, m is the sum of first class index, v_efRepresent that the e two-level index is to f first class index Weighing factor.If and, it is understood that e two-level index and the f first class index are not corresponding relations, then v_efTake Value is 0 (such as: it is then 0 that user watches natural law to the weighing factor of the consuming capacity)；If e two-level index and the f one-level refer to Mark is corresponding relation, then can calculate corresponding value according to the operation of the weighing factor of above-mentioned parameter.

Step S4: according to each two-level index calculated weight W to user's comprehensive value ', use the side of weighted sum Formula calculates comprehensive value score S of each user, defines S_zRepresent the comprehensive value score of z position user, z=1,2 ..., p, p It is represented as maximum number of user amount, proceeds to step S5.

Wherein, S_zComputing formula be: S_z=W₁’*X_z1’+W₂’*X_z2’+…+W_e’*X_ze', X_ze' represent z position user's Value X of e two-level index_zeValue after normalized process, X_ze' computing formula be:

${X_{ze}}^{,}=\frac{X_{ze}-min(X_{1e},X_{2e},...,X_{pe})}{max(X_{1e},X_{2e},...,X_{pe})-\min (X_{1e},X_{2e},...,X_{pe})};$

Max (X in above-mentioned formula_1e, X_2e..., X_pe) represent the maximum of e two-level index in all users；min (X_1e, X_2e..., X_pe) represent and choose the minima of e two-level index in all users.

Step S5: according to comprehensive value score S of each user calculated, chooses corresponding score interval；According to difference Score interval mark off the value grade of each user.In the present embodiment, score interval is divided into from low to high three intervals, Three corresponding basic, normal, high three of score intervals are worth grade.

Shown in Figure 4, the embodiment of the present invention also provides for a kind of user based on step analysis algorithm and is worth marking system. This system includes that analytic hierarchy structure builds module, Index Influence weight computation module, two-level index comprehensive value weight calculation Module, user's comprehensive value computing module, value grade classification module.

Wherein, analytic hierarchy structure builds module and is used for: according to the behavioural information of user, the level building user's value divides Analysis structure, the analytic hierarchy structure that this user is worth is three-decker: ground floor is user's comprehensive value, and the second layer is that user combines Closing and be worth corresponding multiple first class index, third layer is multiple two-level index that each first class index is corresponding；

Index Influence weight computation module is used for: the second layer in user's value level analytical structure is carried out level and divides Analysis, calculates each first class index weighing factor to user's comprehensive value；To the third layer in user's value level analytical structure Carry out step analysis, calculate each two-level index weighing factor to corresponding first class index；

Two-level index comprehensive value weight computation module is used for: according to the impact on user's comprehensive value of each first class index Weight and each two-level index weighing factor to corresponding first class index, calculate each two-level index to user's comprehensive value Weight W ', define W_e' representing the e two-level index weight to user's comprehensive value, e=1,2 ..., k, k is two-level index Sum；

User's comprehensive value computing module is used for: each two calculated according to two-level index comprehensive value weight computation module Level index weight W to user's comprehensive value ', use the mode of weighted sum to calculate comprehensive value score S of each user, fixed Justice S_zRepresenting the comprehensive value score of z position user, z=1,2 ..., p, p is represented as maximum number of user amount；

It is worth grade classification module to be used for: according to the comprehensive value of each user that user's comprehensive value computing module calculates Score S, chooses corresponding score interval；The value grade of each user is marked off according to different score intervals.

It should be understood that the system that above-described embodiment provides is when operating, only drawing with above-mentioned each functional module Divide and be illustrated, in actual application, can as desired above-mentioned functions distribution be completed by different functional modules, i.e. The internal structure of system is divided into different functional modules, to complete all or part of function described above.

The present invention is not limited to above-mentioned embodiment, for those skilled in the art, without departing from On the premise of the principle of the invention, it is also possible to make some improvements and modifications, these improvements and modifications are also considered as the protection of the present invention Within the scope of.

The content not being described in detail in this specification belongs to prior art known to professional and technical personnel in the field.

Claims (10)

1. a user based on step analysis algorithm is worth methods of marking, it is characterised in that the method comprises the following steps:

Step S1: according to the behavioural information of user, build the analytic hierarchy structure that user is worth, the step analysis that this user is worth Structure is three-decker: ground floor is user's comprehensive value, and the second layer is multiple first class index that user's comprehensive value is corresponding, the Three layers is multiple two-level index that each first class index is corresponding, proceeds to step S2；

Step S2: the second layer is carried out step analysis, calculates each first class index weighing factor to user's comprehensive value；To Three layers carry out step analysis, calculate each two-level index weighing factor to corresponding first class index, proceed to step S3；

Step S3: according to each first class index to the weighing factor of user's comprehensive value and each two-level index to corresponding one-level The weighing factor of index, calculates each two-level index weight W to user's comprehensive value ', define W_e' represent that e two grades refer to The mark weight to user's comprehensive value, e=1,2 ..., k, k is the sum of two-level index, proceeds to step S4；

Step S4: according to each two-level index calculated weight W to user's comprehensive value ', use the mode meter of weighted sum Calculate comprehensive value score S of each user, define S_zRepresenting the comprehensive value score of z position user, z=1,2 ..., p, p represents For maximum number of user amount, proceed to step S5；

Step S5: according to comprehensive value score S of each user calculated, chooses corresponding score interval；Obtain according to different By stages marks off the value grade of each user.

2. user based on step analysis algorithm as claimed in claim 1 is worth methods of marking, it is characterised in that: in step S1 The plurality of first class index includes: liveness, the consuming capacity, supplement ability with money；The plurality of two-level index includes: with liveness pair The user answered watches duration, user watches natural law, user watches room number, user sends barrage number, corresponding with the consuming capacity User gives virtual present amount, user gives virtual present room number, the user recharge amount corresponding with the ability of supplementing with money, the most every Secondary recharge amount, supplement natural law with money.

3. user based on step analysis algorithm as claimed in claim 1 is worth methods of marking, it is characterised in that: step S2 In, calculate each first class index to the weighing factor of user's comprehensive value, calculate each two-level index to corresponding first class index The process of weighing factor all includes following operation:

Step 1, selective goal: select the index needing to calculate weighing factor in this step analysis；

Step 2, development of judgment matrix: by the index selected is compared, two-by-two between 1-9 scaling law deliberated index Relative importance grade, construct the judgment matrix A of selected index:

$A=\left[\begin{array}{cccc}1& a_{12}& ...& a_{1n}\\ a_{21}& 1& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& 1\end{array}\right]$

A in definition judgment matrix A_ijRepresent the importance comparative result of index i and index j, and a_ij=1/a_ji, i=1,2 ..., N, j=1,2 ..., n, n represents the number of selected index, the i.e. exponent number of judgment matrix A；

Step 3, calculating weighing factor: current judgment matrix A is pressed row normalization, obtains normalization matrix B, in matrix Bb_ijFor the i-th row normalized value of j row element in matrix A；Normalization matrix B is obtained c by row summation_i,By c_iIt is normalized, obtains the weighing factor w that selected index is corresponding_i, w_iI-th index selected by expression is corresponding Weighing factor, its computing formula is:

Step 4, calculating characteristic vector and eigenvalue of maximum thereof: according to the weighing factor w of currently available selected index_i, worked as The characteristic vector W of front judgment matrix A, W=(w₁, w₂..., w_n)^T, and obtain the eigenvalue of maximum λ that characteristic vector W is corresponding_max,

Step 5, check and adjust judgment matrix, drawing final weighing factor: according to eigenvalue of maximum λ_max, calculate current Coincident indicator CI of judgment matrix A, its computing formula is:Calculate currently according to coincident indicator CI The Consistency Ratio CR of judgment matrix A, its computing formula is: CR=CI/RI, RI are the average homogeneity that judgment matrix exponent number is corresponding Property achievement data；

The Consistency Ratio CR that inspection calculates, whether less than specifying threshold value, if so, shows the discordance degree of judgment matrix A In permissible range, then the most corresponding by obtaining selected index in step 3 weighing factor w_iCorresponding as selected index Whole weighing factor；Otherwise, after adjusting judgment matrix A, repeated execution of steps 3 to step 5, until the CR of judgment matrix A is less than referring to Determine threshold value, then using weighing factor corresponding for the selected index of latest computed as final weighing factor corresponding to selected index.

4. user based on step analysis algorithm as claimed in claim 1 is worth methods of marking, it is characterised in that: step S3 In, W_e' computing formula be:Wherein w_fRepresent the impact on user's comprehensive value of the f first class index Weight, f=1,2 ..., m, m is the sum of first class index, v_efRepresent that the impact of f first class index is weighed by the e two-level index Weight.

5. user based on step analysis algorithm as claimed in claim 1 is worth methods of marking, it is characterised in that: step S4 In, S_zComputing formula be: S_z=W₁’*X_z1’+W₂’*X_z2’+…+W_e’*X_ze', X_ze' represent the e two grades of z position user Refer to target value X_zeValue after normalized process, X_ze' computing formula be:

${X_{ze}}^{,}=\frac{X_{ze}-min(X_{1e},X_{2e},...,X_{pe})}{max(X_{1e},X_{2e},...,X_{pe})-\min (X_{1e},X_{2e},...,X_{pe})};$

Max (X in above-mentioned formula_1e, X_2e..., X_pe) represent the maximum of e two-level index in all users；min(X_1e, X_2e..., X_pe) represent and choose the minima of e two-level index in all users.

6. a user based on step analysis algorithm is worth marking system, it is characterised in that: this system includes that step analysis is tied Structure builds module, Index Influence weight computation module, two-level index comprehensive value weight computation module, the calculating of user's comprehensive value Module, value grade classification module；

Described analytic hierarchy structure builds module and is used for: according to the behavioural information of user, build the step analysis knot that user is worth Structure, the analytic hierarchy structure that this user is worth is three-decker: ground floor is user's comprehensive value, and the second layer is the comprehensive valency of user Multiple first class index that value is corresponding, third layer is multiple two-level index that each first class index is corresponding；

Described Index Influence weight computation module is used for: the second layer in user's value level analytical structure is carried out level and divides Analysis, calculates each first class index weighing factor to user's comprehensive value；To the third layer in user's value level analytical structure Carry out step analysis, calculate each two-level index weighing factor to corresponding first class index；

Described two-level index comprehensive value weight computation module is used for: according to the impact on user's comprehensive value of each first class index Weight and each two-level index weighing factor to corresponding first class index, calculate each two-level index to user's comprehensive value Weight W ', define W_e' representing the e two-level index weight to user's comprehensive value, e=1,2 ..., k, k is two-level index Sum；

Described user's comprehensive value computing module is used for: each two calculated according to two-level index comprehensive value weight computation module Level index weight W to user's comprehensive value ', use the mode of weighted sum to calculate comprehensive value score S of each user, fixed Justice S_zRepresenting the comprehensive value score of z position user, z=1,2 ..., p, p is represented as maximum number of user amount；

Described value grade classification module is used for: according to the comprehensive value of each user that user's comprehensive value computing module calculates Score S, chooses corresponding score interval；The value grade of each user is marked off according to different score intervals.

7. user based on step analysis algorithm as claimed in claim 6 is worth marking system, it is characterised in that: the plurality of First class index includes: liveness, the consuming capacity, supplement ability with money；The plurality of two-level index includes: the user corresponding with liveness Watch duration, user watches natural law, user watches room number, user sends barrage number, and the user corresponding with the consuming capacity gives Virtual present amount, user give virtual present room number, the user recharge amount corresponding with the ability of supplementing with money, averagely supplement gold with money every time Volume, supplement natural law with money.

8. user based on step analysis algorithm as claimed in claim 6 is worth marking system, it is characterised in that: described index Weighing factor computing module calculate each first class index to the weighing factor of user's comprehensive value, calculate each two-level index to phase The process answering the weighing factor of first class index all includes following operation:

Selective goal: select the index needing to calculate weighing factor in this step analysis；

Development of judgment matrix: by the index selected is compared two-by-two, relative according between 1-9 scaling law deliberated index Importance rate, constructs the judgment matrix A of selected index:

$A=\left[\begin{array}{cccc}1& a_{12}& ...& a_{1n}\\ a_{21}& 1& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& 1\end{array}\right]$

A in definition judgment matrix A_ijRepresent the importance comparative result of index i and index j, and a_ij=1/a_ji, i=1,2 ..., N, j=1,2 ..., n, n represents the number of selected index, the i.e. exponent number of judgment matrix A；

Calculate weighing factor: current judgment matrix A is pressed row normalization, obtains normalization matrix B, in matrix B b_ijFor the i-th row normalized value of j row element in matrix A；Normalization matrix B is obtained c by row summation_i,By c_iEnter Row normalization, obtains the weighing factor w that selected index is corresponding_i, w_iThe weighing factor that i-th index selected by expression is corresponding, its meter Calculation formula is:

Calculate characteristic vector and eigenvalue of maximum thereof: according to the weighing factor w of currently available selected index_i, currently judged The characteristic vector W of matrix A, W=(w₁, w₂..., w_n)^T, and obtain the eigenvalue of maximum λ that characteristic vector W is corresponding_max,

Check and adjust judgment matrix, drawing final weighing factor: according to eigenvalue of maximum λ_max, calculate and currently judge square Coincident indicator CI of battle array A, its computing formula is:Calculate according to coincident indicator CI and currently judge square The Consistency Ratio CR of battle array A, its computing formula is: CR=CI/RI, RI are the average homogeneity index that judgment matrix exponent number is corresponding Data；The Consistency Ratio CR that inspection calculates, whether less than specifying threshold value, if so, shows the discordance journey of judgment matrix A Degree is in permissible range, then directly by weighing factor w corresponding for currently available selected index_iAs corresponding final of selected index Weighing factor；Otherwise, after adjusting judgment matrix A, repeated execution of steps 3 to step 5, until the CR of judgment matrix A is less than specifying Threshold value, then using weighing factor corresponding for the selected index of latest computed as final weighing factor corresponding to selected index.

9. user based on step analysis algorithm as claimed in claim 6 is worth marking system, it is characterised in that: described two grades Index comprehensive is worth weight computation module and calculates W_e' formula be:Wherein w_fRepresent the f first class index Weighing factor to user's comprehensive value, f=1,2 ..., m, m is the sum of first class index, v_efRepresent the e two-level index pair The weighing factor of f first class index.

10. user based on step analysis algorithm as claimed in claim 6 is worth marking system, it is characterised in that: described use Family comprehensive value computing module calculates S_zFormula be: S_z=W₁’*X_z1’+W₂’*X_z2’+…+W_e’*X_ze', X_ze' represent z position Value X of e the two-level index of user_zeValue after normalized process, X_ze' computing formula be:

${X_{ze}}^{,}=\frac{X_{ze}-min(X_{1e},X_{2e},...,X_{pe})}{max(X_{1e},X_{2e},...,X_{pe})-\min (X_{1e},X_{2e},...,X_{pe})};$

Max (X in above-mentioned formula_1e, X_2e..., X_pe) represent the maximum of e two-level index in all users；min(X_1e, X_2e..., X_pe) represent and choose the minima of e two-level index in all users.