CN104766215A - Comprehensive multi-dimension goods owner selection quantification method - Google Patents

Abstract

The invention discloses a comprehensive multi-dimension goods owner selection quantification method enabling a member to evaluate a goods owner objectively and comprehensively from multiple dimensions in a quantitative mode. According to the method, multiple measure indexes affecting selection including loading and unloading punctuality, the number of times of carriage, the number of times of complaints from platform members, goods payment and information issuing timeliness ratio, and one-time offer close deal rate are fully considered, the weight of each index is determined quantitatively based on the index weight division method of a multi-attribute decision-making model combining the least square method and the ELECTRE-II method, reasonable evaluation of the credibility of a goods owner on an e-commerce platform with logistics service as transaction object is well achieved through precise big data mining and analyzing and model algorithms, and platform members can judge the comprehensive condition of a goods owner and make choices meeting their own needs.

Description

A kind of owner of cargo that is comprehensive, various dimensions selects quantization method

Technical field

The owner of cargo side's credibility that the present invention relates in a kind of logistics e-commerce platform quantizes and overall evaluation system, is applicable to carry out objective, comprehensive assessment quantification from multiple dimension to the owner of cargo, belongs to data analysis technique field.

Background technology

The development of infotech, has promoted ecommerce and has been combined with the degree of depth of logistic industry, but while a kind of brand-new business model of creation, the transaction of this non-face-to-face is also with and is served potential risk.In the differentiation to risk, detect, in management process, set up a credit security mechanism, allow user to the owner of cargo, the credibility of platform member etc. has one comprehensively to select the suitable owner of cargo, thus saving transportation cost and raising logistic efficiency just seem particularly important.

Existing logistics e-commerce platform indiscriminately imitates the credit assessment method of commodity electron-like business platform mostly, the foundation of assessment indicator system is also relatively unsound, cannot objective, reflect the owner of cargo real load level fully, other users of logistics e-commerce platform cannot be known truly, reliable owner of cargo's load situation.

Summary of the invention

Goal of the invention: for Problems existing in existing logistics e-commerce platform appraisement system with not enough, on all history evaluations in transaction platform and transaction record data basis, the invention provides a kind of owner of cargo side's load level credibility based on combination least square method and Multiple Attribute Decision Model and quantize and comprehensive evaluation system of selection.

Technical scheme: a kind of owner of cargo that is comprehensive, various dimensions selects quantization method, is applicable to provide in the e-commerce platform of physical distribution trading service and carries out Quantitatively Selecting to the owner of cargo.Specifically comprise the steps:

(1) data analysis is chosen confidence level and is comparatively commented scheme attribute, structural scheme collection and property set

The confidence level that reference database and this platform of industry selecting index need comparatively comments scheme attribute, comparatively comments the mutual relationship between scheme and each attribute, set up scheme collection and property set according to confidence level.Confidence level comparatively comments scheme collection: average credibility in of the same trade, commented owner of cargo's confidence level, platform datum confidence level.Property set: choose that consignor freights on time, consignee unloads on time, the owner of cargo carries business number of times, suffer that platform member complains number of times, Information issued promptness rate, owner of cargo's payment for goods pays promptness rate, once offers into single rate seven evaluation indexes as the attribute affecting scheme.

(2) property set standardization

According to attribute, confidence level is comparatively commented to the impact of scheme, the profitable type of attribute of scheme and cost type two class.Its property value of profit evaluation model attribute is the bigger the better, otherwise its property value of cost type attribute is the smaller the better.And the dimension of each scheme attribute and dimensional unit are also different, therefore must carry out nondimensionalization process to each attribute of scheme, the nondimensionalization process of profit evaluation model attribute and cost type attribute is as follows:

Profit evaluation model attribute process: b _ij=(a _ij-a _j ^min)/(a _j ^max-a _j ^min)

Cost type attribute process: b _ij=(a _j ^max-a _ij)/(a _j ^max-a _j ^min)

Wherein, a _ijthe property value of a jth attribute of scheme i, b _ija _ijvalue after standardization, a _j ^maxa jth attribute P _jmaximal value, a _j ^minp _jminimum value.B _ij∈ (0,1), i=1,2 ..., m, j=1,2 ..., n, normalized matrix B=(b _ij) _{m × n}.

(3) carry out multidimensional analysis, build the polynary Optimized model of attribute weight

Consider the validity of information feed back and the rationality of information processing, introduce polynary Optimized model herein, from the dimension that carrier member is different with two, platform, comprehensive consideration they to the Different Effects of owner of cargo's confidence level, and on the basis of large data analysis, mathematics is carried out to them and portrays.Carrier member is different from the contact point of the owner of cargo with platform, therefore shows different attribute evaluation values to the emphasis of owner of cargo's confidence level.

1. polynary weight is determined

In order to reduce single preference and the impact of cognitive limitation on result, the history evaluation marking of all carrier members to each index of the owner of cargo is weighted average by platform, show that carrier member is to owner of cargo's comprehensive evaluation marking the most intuitively; In order to reduce the deviation that single dimension evaluation brings, introduce platform decision-making unit from another dimension herein---based on platform historical trading data and trading activity, supplementary evaluation marking is carried out to owner of cargo's confidence level.Then air exercise is divided into the weighted value that column criterion process obtains each attribute, is finally that two decision-making units arrange significance level coefficient based on the integrality of data and authenticity.

The decision-making unit T that two dimensions are determined _k, k=1,2; Wherein T ₁represent carrier member's decision-making unit, T ₂represent platform decision-making unit.

The attribute weight of two decision-making unit assignment is: w ^k=(w ₁ ^k, w ₂ ^k..., w _n ^k) ^t, k=1,2

The significance level of each decision-making unit is: z=(z ₁, z ₂) ^t, wherein, z ₁+ z ₂=1, z _k>=0.

2. unitary weight optimization model is built

Consider carrier member to the owner of cargo evaluate may with subjective factor, herein from subjective weight determination method angle, comprehensive large data mining technology, builds attribute weight Optimized model as follows:

$\min L_1=\underset{k=1}{\overset{2}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{z}_k{(w_j-{w_j}^k)}^2$

$s.t.\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1,w_j\≥0,j\∈N,$ Wherein N={1,2 ..., n} is property set.

The implication of model tries to achieve a w _j, make w _jand w _j ^kthe quadratic sum L of total partial variance ₁minimum.

3. binary weight optimization model is built

In order to reduce human factor in model as far as possible, herein from objective weight determination method angle, multidimensional analysis is carried out to data, builds attribute weight Optimized model as follows:

G-minL＝(l ₁,l ₂,...l _m)

Wherein, a _j ^*=max{a _1j, a _2j... a _mjbe attribute P _jideal value, with wait power weigthed sums approach, can be by model simplification:

$\min L_2=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{l}_i=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2{w_j}^2$

$s.t.\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1{,w}_j\≥0,j\∈N$

Wherein M={1,2 ..., m} for comparatively to comment scheme collection, N={1,2 ..., n} is the property set of scheme.

4. single object optimization model is synthesized

Above two Optimized models are carried out being integrated into G=min (L ₁, L ₂), and by linear weighting method, problem is converted into following single object optimization model:

$\min F=\frac{1}{2}\underset{k=1}{\overset{2}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{z}_k{(w_j-{w_j}^k)}^2+\frac{1}{2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2{w_j}^2$

$s.t.\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1{,w}_j\≥0,j\∈N$

Wherein, M={1,2 ..., m} for comparatively to comment scheme collection, N={1,2 ..., n} is the property set of scheme.By building Lagrangian function, can obtain:

$w_j=c_j[d_j+(1-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j{d}_j)/\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j]$

Wherein, $c_j=1/[\frac{1}{2}+\frac{1}{2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2],d_j=\frac{1}{2}\underset{k=1}{\overset{q}{\mathrm{\Σ}}}{z}_k{w_j}^k$

(4) attribute weight is solved

Due to multiple decision-making unit (T _k: k=1,2) to the comprehensive weight imparting value of attribute j be:

$w_j=c_j[d_j+(1-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j{d}_j)/\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j],j=\mathrm{1,2},...,n$ For scheme attribute.

The weight vectors of the polynary Optimized model of attribute weight is: w ^*=[w ₁, w ₂... w _j..., w _n], wherein $w_j=c_j[d_j+(1-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j{d}_j)/\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j]$

The final weight matrix of property set is A ^*=(A _ij) _{m × n}=Bw ^{* T}

Wherein, A _ijfor confidence level comparatively comments the final weighted value of the jth of a scheme i attribute, B=(b _ij) _{m × n}for property set normalized matrix.

(5) draft norm confidence score is comparatively commented in calculating, the time m-confidence score curve of record transaction count

Be A according to the final weight matrix of gained property set ^*=(A _ij) _{m × n}=Bw ^{* T}, calculate and comparatively comment draft norm confidence score, and draw the time m-confidence score curve based on transaction count.Main calculation procedure is as follows:

1. calculate the confidence score that confidence level comparatively comments scheme i, formula is

$M_i=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{A}_{\mathrm{ij}}=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{b}_{\mathrm{ij}}{w}_j=[b_{i1},b_{i2},...,b_{\mathrm{in}}]\left[\begin{array}{c}{w}_1\\ w_2\\ \·\·\·\\ w_n\end{array}\right]$

Wherein, b _ijthe property value a of a jth attribute of scheme i _ijvalue after standardization, the weight of each scheme jth attribute drawn after the polynary Optimized model for structure attribute weight.

2. scheme confidence score is comparatively commented in standardization

Confidence level comparatively comments scheme collection to be: (in of the same trade, average credibility, is commented owner of cargo's confidence level, platform datum confidence level), so comparatively comment scheme confidence score collection to be:

M=(M ₁, M ₂, M ₃)=(M _{average credibility in of the same trade}, M _{commented owner of cargo's confidence level}, M _{platform datum confidence level})

Wherein, average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level

With platform datum confidence level M _{platform datum confidence level}for standardized benchmark confidence level is divided, then respectively comparatively comment the standardization confidence score m of scheme _i(i=1,2,3,

Average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level) be:

3. calculate and relative normalizedly comparatively comment scheme confidence score

With standardized platform benchmark confidence level m _{platform datum confidence level}for relative score passes judgment on benchmark letter, then respectively comparatively comment the relative normalized confidence score of scheme

m _i ^*(i＝1,2,3，

Average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level) be:

m ₁ ^*＝m ₁-m ₃

m ₂ ^*＝m ₂-m ₃

m ₃ ^*＝m ₃-m ₃＝0

M ^*=(m ₁ ^*, m ₂ ^*, m ₃ ^*) be and required respectively comparatively comment scheme final confidence score.

Wherein, average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level.

4. the time m-confidence score curve based on transaction count is drawn

Because confidence score can upgrade along with the renewal of platform owner of cargo transaction count, be 0 to remain unchanged according to the known platform datum confidence score of above-mentioned result of calculation, there is practical significance; In of the same trade, average credibility can be constantly updated along with the transaction data accumulation of different industries (or whole platform); Changed by commenting owner of cargo's confidence level to carry out the increase of transaction count along with this owner of cargo at platform.Therefore, take time as transverse axis, confidence score is the longitudinal axis, the time m-confidence score curve based on transaction count can be drawn, interior owner of cargo's confidence level situation of (or whole platform) of the same trade and the confidence level change conditions of certain specific owner of cargo in a period of time can be obtained very intuitively according to curve, helpdesk member better can control risk and select the owner of cargo to carry out order.

(6) relative score of each attribute of concordance exponential sum comparatively commented between scheme is calculated

Platform in order to meet risk partiality or the selection preference of different platform member, by calculating the relative normalized score of concordance exponential sum each attribute, for different platform member provides more diversified and more humane selection.Main calculation procedure is as follows:

1. concordance index is calculated

Concordance index J _ikrefer to the ratio of attribute weight sum shared by the summation of all weights comparatively commenting scheme i not to be inferior to scheme k, according to attribute j (j=1,2, ..., n), if scheme i is better than scheme k (i=1,2 ..., m, k=1,2 ..., m, i ≠ k), be designated as i > k, thus the set of the attribute j of all i of meeting > k is designated as U ⁺(i, k), can obtain U equally ⁼(i, j), specific as follows:

U ⁺(i,k)＝{j\1≤j≤n,A _ij＞A _ik}

U ^＝(i,k)＝{j\1≤j≤n,A _ij＝A _ik}

Then concordance exponential formula is:

$J_{\mathrm{ik}}=\frac{\underset{j\∈U^{+}(i,k)}{\mathrm{\Σ}}{A}_{\mathrm{ij}}+\underset{j\∈U^{=}(i,k)}{\mathrm{\Σ}}{A}_{\mathrm{ij}}}{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{A}_{\mathrm{ij}}}$

2. the relative score of each attribute of the owner of cargo is calculated

Be A according to the final weight matrix of gained property set ^*=(A _ij) _{m × n}=Bw ^{* T}, each attribute of the owner of cargo so relative normalized that to be divided into a _2j ^*=A _2j-A _3j, wherein, j=1,2 ..., n

The relative standard finally obtaining each attribute of the owner of cargo obtains diversity a ₂ ^*=(a ₂₁ ^*, a ₂₂ ^*... a _2n ^*).

Beneficial effect: compared with prior art, the present invention is applicable to provide in the e-commerce platform of physical distribution trading service and carries out trust evaluation to the owner of cargo.Its feature to affect numerous attributes of owner of cargo's confidence level by being divided into the orderly level connected each other, make it methodization, according to the mass data of carrier member and platform, setting platform datum confidence level, the average credibility of certain industry (or whole platform) and the confidence level of certain owner of cargo are carried out with it contrast and draws relative normalized confidence score, result is presented more directly perceived.The confidence level attribute of the owner of cargo is compared between two with platform datum attribute score, then its each attribute is quantitatively described, be convenient to platform member and better select to hold with risk.According to evaluation result, the synthetic reliability situation of the owner of cargo can be known objectively, solve some problem of information asymmetry between client and businessman to a certain extent, make platform member or other members in countless businessman, find the businessman of high credit fast, objectively and make the selection meeting oneself demand.

Accompanying drawing explanation

Fig. 1 is method flow diagram of the present invention;

Fig. 2 be the embodiment of the present invention comparatively comment scheme and attribute model thereof;

Fig. 3 is the time m-confidence score curve based on transaction count.

Embodiment

Below in conjunction with specific embodiment, illustrate the present invention further, these embodiments should be understood only be not used in for illustration of the present invention and limit the scope of the invention, after having read the present invention, the amendment of those skilled in the art to the various equivalent form of value of the present invention has all fallen within the application's claims limited range.

(1) data analysis is chosen confidence level and is comparatively commented scheme attribute, structural scheme collection and property set

The confidence level that reference database and this platform of industry selecting index need comparatively comments scheme attribute, comparatively comments the mutual relationship between scheme and each attribute, set up scheme collection and property set, as Fig. 2 according to confidence level.Confidence level comparatively comments scheme collection: average credibility in of the same trade, commented owner of cargo's confidence level, platform datum confidence level.Property set: choose that consignor freights on time, consignee unloads on time, the owner of cargo carries business number of times, suffer that platform member complains number of times, Information issued promptness rate, owner of cargo's payment for goods pays promptness rate, once offers into single rate seven evaluation indexes as the attribute affecting scheme.The present embodiment comparatively comment the evaluation situation of scheme and each attribute in table 1.

Table 1 embodiment confidence level comparatively comments scheme attribute evaluation table

(2) property set standardization

According to attribute, confidence level is comparatively commented to the impact of scheme, the profitable type of attribute of scheme and cost type two class.Its property value of profit evaluation model attribute is the bigger the better, otherwise its property value of cost type attribute is the smaller the better.And the dimension of each scheme attribute and dimensional unit are also different, therefore must carry out nondimensionalization process to each attribute of scheme, the nondimensionalization process of profit evaluation model attribute and cost type attribute is as follows:

Profit evaluation model attribute process: b _ij=(a _ij-a _j ^min)/(a _j ^max-a _j ^min)

Cost type attribute process: b _ij=(a _j ^max-a _ij)/(a _j ^max-a _j ^min)

Wherein, a _ijthe property value of a jth attribute of scheme i, b _ija _ijvalue after standardization, a _j ^maxa jth attribute P _jmaximal value, a _j ^minp _jminimum value.B _ij∈ (0,1), i=1,2 ..., m, j=1,2 ..., n, normalized matrix B=(b _ij) _{m × n}.

Obtaining confidence level according to table 1 comparatively comments scheme attribute decision matrix as follows:

Wherein, the owner of cargo carries business number of times, Information issued promptness rate, owner of cargo's payment for goods pays promptness rate, once offers into single rate is profit evaluation model attribute, and evaluation of estimate is the bigger the better; Consignor freights on time, consignee unloads on time, suffer platform member to complain number of times is then cost type attribute, and evaluation of estimate is the smaller the better.After carrying out standardization to matrix A, the normalized matrix obtained is:

(3) carry out multidimensional analysis, build the polynary Optimized model of attribute weight

Consider the validity of information feed back and the rationality of information processing, introduce polynary Optimized model herein, from the dimension that carrier member is different with two, platform, comprehensive consideration they to the Different Effects of owner of cargo's confidence level, and on the basis of large data analysis, mathematics is carried out to them and portrays.Carrier member is different from the contact point of the owner of cargo with platform, therefore shows different attribute evaluation values to the emphasis of owner of cargo's confidence level.

Determine polynary weight

In order to reduce single preference and the impact of cognitive limitation on result, platform carries out degree of depth excavation and analysis to carrier member to the large data of magnanimity that the owner of cargo evaluates, and show that carrier member is to owner of cargo's comprehensive evaluation the most intuitively; In order to reduce the deviation that single dimension evaluation brings, introduce platform decision-making unit from another dimension herein---based on platform historical trading data and trading activity, supplementary evaluation is carried out to owner of cargo's confidence level.And be that two decision-making units arrange rational significance level coefficient based on the integrality of data and authenticity.

To carrier member and platform, the marking to each attribute is weighted on average this example, then carries out standardization and obtains:

The attribute weight of carrier member's decision-making unit assignment is: w ¹=(0.53,0.13,0.1,0.6,0.1,0.64,0.64) ^t

The attribute weight of platform decision-making unit assignment is: w ²=(0.2,0.2,0.4,0.75,0.15,0.9,0.4) ^t

The significance level of each decision-making unit is: z=(0.45,0.55) ^t.

(4) attribute weight is solved

In order to make attribute weight both comprise subjective preferences, again containing objective information, subjective weight optimization model and objective weight Optimized model can be carried out being integrated into G=min (L ₁, L ₂), and by linear weighting method, problem is converted into following single object optimization model:

$\min F=\frac{1}{2}\underset{k=1}{\overset{2}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{z}_k{(w_j-{w_j}^k)}^2+\frac{1}{2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2{w_j}^2$

$s.t.\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1{,w}_j\≥0,i\∈M,j\∈N$

Wherein, M={1,2 ..., m} for comparatively to comment scheme collection, N={1,2 ..., n} is the property set of scheme.

By building Lagrangian function, can obtain:

c＝(0.89,0.89,0.33,0.67,0.4,0.22,0.1) ^T

d＝(0.15,0.1,0.11,0.31,0.06,0.38,0.25) ^T

According to formula, then can obtain and reflect that the weight vectors of carrier member and platform information is simultaneously:

w ^*＝(0.23,0.19,0.07,0.28,0.07,0.11,0.04) ^T

(5) draft norm confidence score is comparatively commented in calculating, the time m-confidence score curve of record transaction count

Be A according to the final weight matrix of gained property set ^*=(A _ij) _{m × n}=Bw ^{* T}, calculate and comparatively comment draft norm confidence score, and draw the time m-confidence score curve based on transaction count.Main calculation procedure is as follows:

Calculate the confidence score that confidence level comparatively comments scheme i, formula is

$M_i=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{A}_{\mathrm{ij}}=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{b}_{\mathrm{ij}}{w}_j=[b_{i1},b_{i2},...,b_{\mathrm{in}}]\left[\begin{array}{c}{w}_1\\ w_2\\ \·\·\·\\ w_n\end{array}\right]$

Confidence level comparatively comments scheme collection to be: (in of the same trade, average credibility, is commented owner of cargo's confidence level, platform datum confidence level), so comparatively comment scheme confidence score collection to be:

M=(M ₁, M ₂, M ₃)=(M _{average credibility in of the same trade}, M _{commented owner of cargo's confidence level}, M _{platform datum confidence level})

Wherein, average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level

Try to achieve according to formula:

M=(M ₁, M ₂, M ₃)=(M _{average credibility in of the same trade}, M _{commented owner of cargo's confidence level}, M _{platform datum confidence level})=(0.5,0.75,0.28)

Scheme confidence score is comparatively commented in standardization

With platform datum confidence level M _{platform datum confidence level}for standardized benchmark confidence level is divided, then respectively comparatively comment the standardization confidence score m of scheme _i(i=1,2,3,

Average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level) be:

Institute is in the hope of m=(1.76,2.64,1)

Calculate and relative normalizedly comparatively comment scheme confidence score

With standardized platform benchmark confidence level m _{platform datum confidence level}for relative score passes judgment on benchmark letter, then respectively comparatively comment the relative normalized confidence score of scheme

m _i ^*(i＝1,2,3，

Average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level) be:

m ₁ ^*＝m ₁-m ₃

m ₂ ^*＝m ₂-m ₃

m ₃ ^*＝m ₃-m ₃＝0

Obtain m ^*=(m ₁ ^*, m ₂ ^*, m ₃ ^*)=(0.76,1.64,0) be and requiredly respectively comparatively comment scheme final confidence score.Wherein, average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level.

As shown in Figure 3, the time m-confidence score curve based on transaction count is drawn

Because confidence score can upgrade along with the renewal of platform owner of cargo transaction count, be 0 to remain unchanged according to the known platform datum confidence score of above-mentioned result of calculation, there is practical significance; In of the same trade, average credibility can be constantly updated along with the transaction data accumulation of different industries (or whole platform); Changed by commenting owner of cargo's confidence level to carry out the increase of transaction count along with this owner of cargo at platform.Therefore, take time as transverse axis, confidence score is the longitudinal axis, the time m-confidence score curve based on transaction count can be drawn, interior owner of cargo's confidence level situation of (or whole platform) of the same trade and the confidence level change conditions of certain specific owner of cargo in a period of time can be obtained very intuitively according to curve, helpdesk member better can control risk and select the owner of cargo to carry out order.

(6) relative score of each attribute of the owner of cargo is calculated

Be A according to the final weight matrix of gained property set ^*=(A _ij) _{m × n}=Bw ^{* T}, each attribute of the owner of cargo so relative normalized that to be divided into a _2j ^*=A _2j-A _3j, wherein, j=1,2, ..., 7,1=consignor freights on time, 2=consignee unloads on time, the 3=owner of cargo carries business number of times, 4=suffers platform member to complain number of times, 5=Information issued promptness rate, 6=owner of cargo's payment for goods payment promptness rate, 7=once to offer into single rate.

The relative standard finally obtaining each attribute of the owner of cargo must be divided into:

a ₂ ^*＝(a ₂₁ ^*,a ₂₂ ^*,...a _2n ^*)＝(0.2,0.1,0.04,0,0.07,0,0.04)

Wherein, a ₂₄ ^*=0, a ₂₆ ^*=0 to show to be commented the owner of cargo " to suffer platform member to complain number of times " identical with platform datum score with " owner of cargo's payment for goods payment promptness rate " these two attribute scores, be that 0 point of explanation is performed poor on this attribute relative to the datum-plane owner of cargo herein, just just pass, when selecting, platform member should notice that this owner of cargo exists higher confidence level risk in these two, should be careful.

Claims (4)

1. the owner of cargo that is comprehensive, various dimensions selects a quantization method, it is characterized in that, specifically comprises the steps:

(1) data analysis is chosen confidence level and is comparatively commented scheme attribute, structural scheme collection and property set

(2) property set standardization

Carry out nondimensionalization process to each attribute of scheme, the nondimensionalization process of profit evaluation model attribute and cost type attribute is as follows:

Profit evaluation model attribute process: b _ij=(a _ij-a _j ^min)/(a _j ^max-a _j ^min)

Cost type attribute process: b _ij=(a _j ^max-a _ij)/(a _j ^max-a _j ^min)

Wherein, a _ijthe property value of a jth attribute of scheme i, b _ija _ijvalue after standardization, a _j ^maxa jth attribute P _jmaximal value, a _j ^minp _jminimum value.B _ij∈ (0,1), i=1,2 ..., m, j=1,2 ..., n, normalized matrix B=(b _ij) _{m × n};

(3) carry out multidimensional analysis, build the polynary Optimized model of attribute weight

Introduce polynary Optimized model, from the dimension that carrier member is different with two, platform, comprehensive consideration they to the Different Effects of owner of cargo's confidence level, and on the basis of large data analysis, mathematics is carried out to them and portrays; Carrier member is different from the contact point of the owner of cargo with platform, therefore shows different attribute evaluation values to the emphasis of owner of cargo's confidence level;

(4) attribute weight is solved

Due to multiple decision-making unit (T _k: k=1,2) to the comprehensive weight imparting value of attribute j be:

$w_j=c_j[d_j+(1-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j{d}_j)/\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j],j=\mathrm{1,2},...,n$

The weight vectors of the polynary Optimized model of attribute weight is: w ^*=[w ₁, w ₂... w _j..., w _n], wherein $w_j=c_j[d_j+(1-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j{d}_j)/\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j]1$

The final weight matrix of property set is A ^*=(A _ij) _{m × n}=Bw ^{* T}

Wherein, A _ijfor confidence level comparatively comments the final weighted value of the jth of a scheme i attribute, B=(b _ij) _{m × n}for property set normalized matrix;

(5) draft norm confidence score is comparatively commented in calculating, the time m-confidence score curve of record transaction count

Be A according to the final weight matrix of gained property set ^*=(A _ij) _{m × n}=Bw ^{* T}, calculate and comparatively comment draft norm confidence score, and draw the time m-confidence score curve based on transaction count;

(6) relative score of each attribute of concordance exponential sum comparatively commented between scheme is calculated

Platform in order to meet risk partiality or the selection preference of different platform member, by calculating the relative normalized score of concordance exponential sum each attribute, for different platform member provides more diversified and more humane selection.

2. the owner of cargo that is comprehensive, various dimensions selects quantization method as claimed in claim 1, it is characterized in that, the concrete steps of step 3 are:

1. polynary weight is determined

In order to reduce single preference and the impact of cognitive limitation on result, the history evaluation marking of all carrier members to each index of the owner of cargo is weighted average by platform, show that carrier member is to owner of cargo's comprehensive evaluation marking the most intuitively; In order to reduce the deviation that single dimension evaluation brings, introduce platform decision-making unit from another dimension herein---based on platform historical trading data and trading activity, supplementary evaluation marking is carried out to owner of cargo's confidence level; Then air exercise is divided into the weighted value that column criterion process obtains each attribute, is finally that two decision-making units arrange significance level coefficient based on the integrality of data and authenticity;

The decision-making unit T that two dimensions are determined _k, k=1,2; Wherein T ₁represent carrier member's decision-making unit, T ₂represent platform decision-making unit;

The attribute weight of two decision-making unit assignment is: w ^k=(w ₁ ^k, w ₂ ^k..., w _n ^k) ^t, k=1,2

The significance level of each decision-making unit is: z=(z ₁, z ₂) ^t, wherein, z ₁+ z ₂=1, z _k>=0;

2. unitary weight optimization model is built

Consider carrier member to the owner of cargo evaluate may with subjective factor, herein from subjective weight determination method angle, comprehensive large data mining technology, builds attribute weight Optimized model as follows:

$\min L_1=\underset{k=1}{\overset{2}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{z}_k{(w_j-{w_j}^2)}^2$

$\begin{array}{ccc}s.t.& \underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1& w_j\≥0,j\∈N\end{array}$

Wherein N={1,2 ..., n} is the property set of scheme

The implication of model tries to achieve a w _j, make w _jand w _j ^kthe quadratic sum L of total partial variance ₁minimum;

3. binary weight optimization model is built

Herein from objective weight determination method angle, multidimensional analysis is carried out to data, builds attribute weight Optimized model as follows:

G-minL＝(l ₁,l ₂,...l _m)

Wherein, $l_i=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2{w_j}^2,{a_j}^{*}=\max \{a_{1j},a_{2j},\·\·\·a_{\mathrm{mj}}\}$ For attribute P _jideal value, with wait power weigthed sums approach, can be by model simplification:

$\min L_2=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{l}_i=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2{w_j}^2$

$\begin{array}{ccc}s.t.& \underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1& w_j\≥0,i\∈M,j\∈N\end{array}$

Wherein M={1,2 ..., m} for comparatively to comment scheme collection, N={1,2 ..., n} is the property set of scheme

4. single object optimization model is synthesized

Above two Optimized models are carried out being integrated into G=min (L ₁, L ₂), and by linear weighting method, problem is converted into following single object optimization model:

$\min F=\frac{1}{2}\underset{k=1}{\overset{2}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{z}_k{(w_j-{w_j}^k)}^2+\frac{1}{1}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2{w_j}^2$

$\begin{array}{ccc}s.t.& \underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1& w_j\≥0,i\∈M,j\∈N\end{array}$

Wherein, M={1,2 ..., m} for comparatively to comment scheme collection, N={1,2 ..., n} is the property set of scheme.By building Lagrangian function, can obtain:

$w_j=c_j[d_j+(1-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j{d}_j)/\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{c}_j]$

Wherein, $c_j=1/[\frac{1}{2}+\frac{1}{2}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({a_j}^{*}-a_{\mathrm{ij}})}^2,d_j=\frac{1}{2}\underset{k=1}{\overset{2}{\mathrm{\Σ}}}{z}_k{w_j}^k.$

3. the owner of cargo that is comprehensive, various dimensions selects quantization method as claimed in claim 2, it is characterized in that, the main calculation procedure of step (5) is as follows:

1. calculate the confidence score that confidence level comparatively comments scheme i, formula is

$M_i=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{A}_{\mathrm{ij}}=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}{w}_j=[b_{i1},b_{i2},...,b_{\mathrm{in}}]\left[\begin{array}{c}{w}_1\\ w_2\\ ...\\ w_n\end{array}\right]$

Wherein, b _ijthe property value a of a jth attribute of scheme i _ijvalue after standardization, the weight of each scheme jth attribute drawn after the polynary Optimized model for structure attribute weight;

2. scheme confidence score is comparatively commented in standardization

Confidence level comparatively comments scheme collection to be: (in of the same trade, average credibility, is commented owner of cargo's confidence level, platform datum confidence level), so comparatively comment scheme confidence score collection to be:

M=(M ₁, M ₂, M ₃)=(M _{average credibility in of the same trade}, M _{commented owner of cargo's confidence level}, M _{platform datum confidence level})

Wherein, average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level

With platform datum confidence level M _{platform datum confidence level}for standardized benchmark confidence level is divided, then respectively comparatively comment the standardization confidence score m of scheme _i(i=1,2,3,

Average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level) be:

3. calculate and relative normalizedly comparatively comment scheme confidence score

With standardized platform benchmark confidence level m _{platform datum confidence level}for relative score passes judgment on benchmark letter, then respectively comparatively comment the relative normalized confidence score of scheme

m _i ^*(i＝1,2,3，

Average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level) be:

m ₁ ^*＝m ₁-m ₃

m ₂ ^*＝m ₂-m ₃

m ₃ ^*＝m ₃-m ₃＝0

M ^*=(m ₁ ^*, m ₂ ^*, m ₃ ^*) be and required respectively comparatively comment scheme final confidence score;

Wherein, average credibility in the 1=same industry, 2=is commented owner of cargo's confidence level, 3=platform datum confidence level;

4. the time m-confidence score curve based on transaction count is drawn

Because confidence score can upgrade along with the renewal of platform owner of cargo transaction count, be 0 to remain unchanged according to the known platform datum confidence score of above-mentioned result of calculation, there is practical significance; In of the same trade, average credibility can be constantly updated along with the transaction data accumulation of different industries (or whole platform); Changed by commenting owner of cargo's confidence level to carry out the increase of transaction count along with this owner of cargo at platform.Therefore, take time as transverse axis, confidence score is the longitudinal axis, the time m-confidence score curve based on transaction count can be drawn, interior owner of cargo's confidence level situation of (or whole platform) of the same trade and the confidence level change conditions of certain specific owner of cargo in a period of time can be obtained very intuitively according to curve, helpdesk member better can control risk and select the owner of cargo to carry out order.

4. the owner of cargo that is comprehensive, various dimensions selects quantization method as claimed in claim 1, it is characterized in that, the main calculation procedure of step (6) is as follows:

1. concordance index is calculated

Concordance index J _ikrefer to the ratio of attribute weight sum shared by the summation of all weights comparatively commenting scheme i not to be inferior to scheme k, foundation attribute j (j=1,2 ..., n), if scheme i be better than scheme k (i=1,2 ..., m, k=1,2 ..., m, i ≠ k), be designated as thus meet all the set of attribute j be designated as U ⁺(i, k), can obtain U equally ⁼(i, k), specific as follows:

Then concordance exponential formula is:

$J_{\mathrm{ik}}=\frac{\underset{j\∈U^{+}(i,k)}{\mathrm{\Σ}}{A}_{\mathrm{ij}}+\underset{j\∈U^{=}\left(\mathrm{imk}\right)}{\mathrm{\Σ}}{A}_{\mathrm{ij}}}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_{\mathrm{ij}}}$

2. the relative score of each attribute of the owner of cargo is calculated

Be A according to the final weight matrix of gained property set ^*=(A _ij) _{m × n}=Bw ^{* T}, each attribute of the owner of cargo so relative normalized that to be divided into a _2j ^*=A _2j-A _3j, wherein, j=1,2 ..., n is scheme attribute;

The relative standard finally obtaining each attribute of the owner of cargo obtains diversity a ₂ ^*=(a ₂₁ ^*, a ₂₂ ^*... a _2n ^*).