CN108596488A - One kind is based on entropy standardization locality preserving projections city logistics level evaluation method - Google Patents

One kind is based on entropy standardization locality preserving projections city logistics level evaluation method Download PDF

Info

Publication number
CN108596488A
CN108596488A CN201810376569.6A CN201810376569A CN108596488A CN 108596488 A CN108596488 A CN 108596488A CN 201810376569 A CN201810376569 A CN 201810376569A CN 108596488 A CN108596488 A CN 108596488A
Authority
CN
China
Prior art keywords
pca
standardization
city
elpp
index
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201810376569.6A
Other languages
Chinese (zh)
Inventor
陶新民
刘锐
梁祥磊
任超
郭文杰
刘艳超
陶思睿
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeast Forestry University
Original Assignee
Northeast Forestry University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northeast Forestry University filed Critical Northeast Forestry University
Priority to CN201810376569.6A priority Critical patent/CN108596488A/en
Publication of CN108596488A publication Critical patent/CN108596488A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/083—Shipping

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Marketing (AREA)
  • Educational Administration (AREA)
  • Tourism & Hospitality (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

One kind is based on entropy standardization locality preserving projections city logistics level evaluation method, it is characterised in that：This approach includes the following steps：(1) it determines city logistics assessment of levels criterion, and carries out the classification of output and input-occupancy-output analysis；(2) input-occupancy-output analysis obtained to step (1) is inverted and all indexs are normalized using minimax method, makes its range between [0,1]；(3) the normalization data matrix X that step (2) is obtained*Pivot constituent analysis and entropy standardization locality preserving projections analysis are carried out, it is final to determine projection vector PPcaAnd WELpp；(4) projection vector that step (3) obtains is utilized to calculate global distribution characteristics coefficient vector X 'Pca=PPca TX*And local distribution characteristic coefficient vector X 'ELpp=WELpp TX*；(5) X ' to being acquired in step (6)PcaWith X 'ELppIt is weighted processing and obtains final evaluation coefficient vector X ', be final city logistics assessment of levels result to the arrangement of X ' carry out descendings.The present patent application is applied to the evaluation problem of city logistics level.

Description

One kind is based on entropy standardization locality preserving projections city logistics level evaluation method
Technical field：
The present application relates to a kind of logistics level evaluation field more particularly to a kind of entropy standardization locality preserving projections point The method of the city logistics assessment of levels of analysis.
Background technology：
Not only there is effect outstanding in city in politics and culture life, but also has important economic status, is commodity The production center, collecting and distributing centre and the consumer center, have stronger comprehensive service capability, radianting capacity and attraction power.And city City's logistics is the important component of Comprehensive Competitiveness of Cities, plays important facilitation to the development of urban economy, therefore City logistics development level is scientifically and rationally evaluated, logistics system can be both optimized, the logistics service for improving city is horizontal, promotes Commodity circulation, it helps promote the whole competitiveness in city.City logistics refer to all logistics activities in urban area, Include not only in the flowings and logistics progress of functional entitys such as storage, packaging, transport, handling, circulation and process and information processing The item motion of each link, further include city and perimeter cargo is collecting and distributing or even the process of urban waste cleaning.City Logistics has following characteristics：(1) city logistics are not only present in production field, the field of circulation, and are also present in consumption neck Domain.(2) material goods of city logistics flowing not only have the productions such as the required various raw material of productive consumption, machinery equipment Data, while also having the various consumer goods of people's personal consumption, or even including urban waste article.(3) management of city logistics relates to And wide, thousands of a families and business unit in existing city, and have the family and enterprise in other cities and surrounding rural area Unit.
In the evaluation method of city logistics development level, more commonly used method has：Linear weight method, step analysis Method, ABC cost-or-markets method, data envelope analysis, Principal Component Analysis etc., wherein Principal Component Analysis are in supplier evaluation using the most Extensively.Principal component analysis (Principal Components Analysis, PCA) is exactly to refer to multiple index comprehensives at minority A kind of multivariate statistical method of target remains original index in such a way that original index is linearly combined into overall target Main information and uncorrelated to each other, greatly reduces the difficulty of overall merit.However traditional principal component analysis can only remove It is linearly related between index, when between the indices of city logistics assessment of levels in practice there are when non-linear relation, traditional master Constituent analysis is unable to get satisfied result.In addition principal component analysis only takes into account global distribution characteristics, that is, only takes into account projection Parameter variance afterwards maximizes, and could not be fully considered to the holding of local message, therefore the change by being obtained after PCA dimensionality reductions Although amount can reflect most global informations of original variable, the local neighborhood structure of legacy data can be upset, and then lead Cause the forfeiture of microcosmic evaluating ability.
Locality preserving projections (Locality Preserving Projections, LPP), are nonlinear methods The linear approximation of Laplacian Eigenmap, as a kind of new subspace analysis method, it both solves principal component analysis Method is difficult to keep the problem of initial data non-linearity manifold, and solves it and can not remove asking for non-linear dependencies between index Topic.LPP is used widely in fields such as recognition of face, image retrievals now.However LPP need to specify in advance k close to number with And the parameter of heat kernel function determines similarity matrix, and in practical application, due in advance to the spatial distribution of initial data Unknown, parameter setting becomes very difficult.It is influenced by improper parameter setting, the performance degradation of traditional LPP algorithms.For This, the present invention standardizes by using entropy is attached to similarity matrix in majorized function, is solved together with projection vector, in turn It solves the deficiency that traditional LPP algorithms need to specify parameter in advance, improves the local space holding capacity of algorithm.The present invention plans The locality preserving projections analysis of entropy standardization and principal component analytical method combine to realize city logistics assessment of levels, both have The global information ability of maintenance of PCA, i.e. macroscopic evaluation ability, and maintain the local relation of former data structure, i.e., microcosmic evaluation Ability.
Invention content：
It is an object of the present invention to provide a kind of city logistics level evaluation methods based on entropy standardization locality preserving projections.
Above-mentioned purpose is realized by the following technical solution：
1, a kind of based on entropy standardization locality preserving projections city logistics level evaluation method.It is characterized in that：This method Include the following steps：
(1) related city is determined by government statistics yearbook or in the way of with related field expert exchanging interview and questionnaire survey Logistics level interpretational criteria, if N number of evaluation index altogether；
(2) evaluation index obtained to step (1) is classified according to output index and input-occupancy-output analysis, if shared L production Go out evaluation index and P input evaluation index, wherein N=L+P；
(3) the classification indicators information obtained according to step (2), collects the various indexs in relation to city logistics assessment of levels Value is equipped with M city, X ∈ RN×M；
(4) the city logistics assessment of levels index value obtained to step (3) carries out numeralization processing, generates sample matrix, The inverted processing of input-occupancy-output analysis information in city will be studied, then each index value is standardized, makes each index Numberical range determine between [0,1]；
(5) the sample matrix X after the horizontal index value standardization of the city logistics for the participation evaluation for obtaining step (4)*∈RN ×M, carry out entropy standardization locality preserving projections analysis and principal component analysis, wherein principal component analysis retain pivot number and The projection vector number of locality preserving projections is respectively 1.It is final to determine PPcaAnd WELppProjection vector, PPca, WELpp∈RN×1；
(6) the projection vector P obtained in step (5) is utilizedPcaSolve global distribution characteristics coefficient vector X 'Pca=PPca TX* And local distribution characteristic coefficient vector X 'ELpp=WELpp TX*, X 'Pca, X 'ELpp∈R1×M；
(7) the global distribution characteristics coefficient vector X ' to being acquired in step (6)PcaDrawn game portion distribution characteristics coefficient vector X′ELppIt is weighted processing and obtains final evaluation coefficient vector：X '=X 'Pca+βX′ELpp, X ' ∈ R1×M.It is recommended that β=1, right X ' carry out descending arrangements, as final N number of city logistics assessment of levels result.
It is 2, according to claim 1 to be based on entropy standardization locality preserving projections city logistics level evaluation method, It is characterized in that, output index refers to that can reflect that city logistics infrastructure situation, development of logistics line are horizontal, believe in step (2) The index of breathization degree and the level of economic development, export-oriented degree and the level of consumption, value show more greatly the level of city logistics It is stronger.And input-occupancy-output analysis refers to city logistics development negative effect caused by society, logistic industry pollutant emission level is handed over Logical stopping state, vehicle, steamer and aircraft noise level, the vibrations of lorry and the damaged condition of road pavement etc., value is smaller Better.
It is 3, according to claim 1 to be based on entropy standardization locality preserving projections city logistics level evaluation method, It is characterized in that, to the inverted processing of input-occupancy-output analysis in step (4), xij∈ X, i=1,2 ..., P, j=1,2 ..., M, xij= 1/xij.Method is maximin method used by being standardized to each index value, is as follows：If xij∈ X, i=1,2 ..., N, j=1,2 ..., M, to arbitrary index xijMethod is as follows used by being standardized：
It is 4, according to claim 1 to be based on entropy standardization locality preserving projections city logistics level evaluation method, It is characterized in that, P in step (5)PcaComputational methods it is as described below：
To the sample matrix X after standardization*Carry out equalization processing：
Wherein
Obtain equalization treated sample matrixMaximum global characteristics can be kept in order to ask, that is, meet projection coefficient The projection vector P of maximum variancePca, enableCalculate characteristic equation：∑PPca=λ PPca, since ∑ is real symmetrical square Battle array, therefore the feature vector corresponding to its maximum eigenvalue can be acquired i.e.：PPca。
It is 5, according to claim 1 to be based on entropy standardization locality preserving projections city logistics level evaluation method, It is characterized in that, W in step (5)ELppComputational methods are as described below：
According to the sample matrix X after standardization*Construct adjacency matrix U(1)∈RM×MWith metric matrix D(1)∈RM×M, wherein uij (1)∈U(1)Expression formula it is as follows：
xi=[x1i, x2i..., xNi]T, xj=[x1j, x2j..., xNj]T, i, j=1,2 ..., M, because of the side in the present invention Method is insensitive to the selection of σ and K, it is therefore proposed that σ=1 or 0.5, are arranged K=5.
Construct metric matrix D(1), dij (1)=0, i ≠ j,I, j=1,2..., M.And it acquires Laplacian Matrix L(1), wherein L(1)=D(1)-U(1)。
Construct A=XL(1)XT, B=XD(1)XTAnd AV=λ BV generalized eigenvalue feature vectors are solved, wherein minimum feature Feature vector V corresponding to value λ is
Maximum cycle MaxIter=300 and current iteration number iter=2 is set, as iter≤MaxIter When
Reconfigure U(iter), according to U(iter)Construct metric matrix D(iter),dij (iter)=0, i ≠ j,I, j=1,2..., M and L(iter), construct A=XL(iter)XT, B=XD(iter)XTAnd AV=λ BV generalized eigenvalue feature vectors are solved, wherein the feature vector V corresponding to minimum eigenvalue λ is Iter=iter+1 is repeated until meeting stop condition.WELppIt is equal to finally obtainI.e.
It is 6, according to claim 1 to be based on entropy standardization locality preserving projections city logistics level evaluation method, It is characterized in that, the projection vector P that step (5) acquiresPcaThe global distributed architecture in luv space can farthest be retained, i.e., Maximum variance after projection is conducive to the performance evaluation comparison in macro-scope；Projection vector W in step (5)ELppIt can be maximum Retain to degree the local neighborhood structure in luv space, that is, the local neighborhood after projecting remains unchanged, and is conducive to microscopic ranges Interior performance evaluation comparison.
Beneficial effects of the present invention：
1. it is a kind of by indices quantitative analysis the city logistics level evaluation method of the present invention, to comprehensively The method that objectively city logistics level is evaluated.The index wherein used is divided into output index and input-occupancy-output analysis, without It is only with a kind of index so that vendors' evaluating result is more comprehensive, true, objective.
2. the city logistics level evaluation method of the present invention, using entropy standardization locality preserving projections and principal component analysis phase In conjunction with method and non-traditional principal component analytical method realize coordinate projection transformation.Though traditional principal component analytical method can be maximum The global structure information of initial data is kept to degree, but the nonlinear correlation problem between evaluation index and holding office can not be handled Portion's structural information, and nonlinear correlation relationship is not limited the method for locality preserving projections analysis between by index, and with part Structure holding capacity.Therefore two kinds of algorithms are combined by the present invention, give full play to the global structure holding capacity of principal component analysis The local message holding capacity of Projection Analysis is kept with part so that method of the invention can also can be from microcosmic from macroscopic perspective Angle realizes the evaluation analysis of city logistics level.
3. the city logistics level evaluation method of the present invention, it is proposed that a kind of locality preserving projections point based on entropy standardization Similarity matrix and projection vector are optimized by using maximum entropy standardization item, solve traditional LPP algorithms by analysis method jointly The deficiency that parameter need to be specified in advance greatly improves algorithm part holding capacity and has higher stability.
Description of the drawings：
Attached drawing 1 be the present invention implementation 2 under wine data sets LPP algorithms 2 dimension dimensionality reduction projection coefficient distribution maps.
Attached drawing 2 be the present invention implementation 2 in 2 dimensions of entropy standardization LPP algorithms after the first iteration under wine data sets Dimensionality reduction projection coefficient distribution map.
Attached drawing 3 be the present invention implementation 2 in 2 dimensions of the entropy standardization LPP algorithms after the 300th iteration under wine data sets Dimensionality reduction projection coefficient distribution map.
Attached drawing 4 be the present invention implementation 2 under wine data sets PCA algorithms 2 dimension dimensionality reduction projection coefficient distribution maps.
Attached drawing 5 be the present invention implementation 2 in entropy standardization LPP 300 iterative target functional values of algorithm under wine data sets Variation diagram.
Attached drawing 6 be the present invention implementation 2 under wine data sets tradition LPP algorithms projected in arest neighbors number k=1 Index profile.
Attached drawing 7 be the present invention implementation 2 under wine data sets tradition LPP algorithms projected in arest neighbors number k=5 Index profile.
Attached drawing 8 be the present invention implementation 2 under wine data sets tradition LPP algorithms projected in arest neighbors number k=10 Index profile.
Attached drawing 9 be the present invention implementation 2 under wine data sets tradition LPP algorithms projected in arest neighbors number k=15 Index profile.
Entropy standardization LPP algorithms are projected in arest neighbors number k=1 under wine data sets in the implementation 2 of Figure 10 present invention Index profile.
Entropy standardization LPP algorithms are projected in arest neighbors number k=5 under wine data sets in the implementation 2 of Figure 11 present invention Index profile.
Entropy standardization LPP algorithms are thrown in arest neighbors number k=10 under wine data sets in the implementation 2 of Figure 12 present invention Shadow index profile.
Entropy standardization LPP algorithms are thrown in arest neighbors number k=15 under wine data sets in the implementation 2 of Figure 13 present invention Shadow index profile.
The logistics level evaluation index value of Shandong Province main cities in the implementation 3 of Figure 14 present invention.
The logistics level of Shandong Province main cities (containing two virtual cities) evaluates max- in the implementation 3 of Figure 15 present invention Index value after min standardization.
The logistics level evaluation index autocorrelation matrix of Shandong Province main cities in the implementation 3 of Figure 16 present invention.
The logistics level evaluation result of Shandong Province main cities (containing two virtual cities) in the implementation 3 of Figure 17 present invention.
The logistics level evaluation index value of Jilin Province main cities in the implementation 3 of Figure 18 present invention.
The logistics level of Jilin Province main cities (containing two virtual cities) evaluates max- in the implementation 3 of Figure 19 present invention Index value after min standardization.
The logistics level evaluation index autocorrelation matrix of Jilin Province main cities in the implementation 3 of Figure 20 present invention.
The logistics level evaluation result of Jilin Province main cities (containing two virtual cities) in the implementation 3 of Figure 21 present invention.
Specific implementation mode：
Embodiment 1：
One kind is described based on entropy standardization office based on entropy standardization locality preserving projections city logistics level evaluation method Portion keeps analysis method to comprise the following specific steps that：
Given data collection x1, x2... xM, wherein xi∈RN, i=1,2 ..., M.M is total sample number, and N is index sum.If yi,yjIt is the One Dimensional Projection coordinate under new base vector, considers in population sample, originally similar sample xi,xjIn new base It marks also the same close in space.The similarity matrix is enabled to be：U∈RM×M.Wherein uij∈ U indicate the sample x of original sample spacei,xj Close degree, i, j=1,2,3..., M, metric matrix：D∈RM×M, wherein dij=∑j uij, utilize maximum entropy standardization structure It makes and solves U and projection vector WELppObject function be：
Wherein α indicates the control weight of regulation loss item and maximum entropy regular terms, enables α=1 here.
Therefore Lagrangian is constructed using lagrange multiplier approach, formula is as follows：
Respectively to β, uijDerivation obtains：
U is acquired by above formulaijExpression formula,
The u that will be acquiredijExpression formula be brought intoIn
Formula is updated to uijIn expression formula, u is obtainedijIterative formula
For the ease of to WELppProjection vector derivation, we change the loss function in above-mentioned object function into matrix Expression-form：
L=(D-U) is famous Laplacian Matrix.0 value solution in order to prevent, in addition constraints：Write as matrix form i.e.：
WELpp TXDXTWELpp=1
W is solved by LagrangeELpp, obtained projection vector iterative formula is expressed as
2XLXTWELpp-2λXXTWELpp=0
XLXTWELpp=λ XXTWELpp
The generalized eigenvalue and feature vector of solution above formula are to get WELpp。
Specific iterative process is as follows：
(1) the adjacency matrix U of Weight is constructed(1)∈RM×MWith metric matrix D(1)∈RM×M, wherein uij (1)∈U(1)Table It is as follows up to formula：
xi=[x1i, x2i..., xNi]T, xj=[x1j, x2j..., xNj]T, i, j=1,2 ..., M, because of the side in the present invention Method is insensitive to the selection of σ and K, it is therefore proposed that σ=1 or 0.5, are arranged K=5.
Construct metric matrix D(1), dii (1)∈D(1), dij (1)=0, i ≠ j,I, j=1, 2..., M.And acquire Laplacian Matrix L(1), wherein L(1)=D(1)-U(1).Construct A=XL(1)XT, B=XD(1)XTAnd solve AV =λ BV generalized eigenvalue feature vectors, wherein the feature vector V corresponding to minimum eigenvalue λ is
(2) maximum cycle Maxlter=300 and current iteration number iter=2 is set.
(3) as iter≤Maxlter
Reconfigure U(iter), according to U(iter)Construct metric matrix D(iter),dij (iter)=0, i ≠ j,I, j=1,2..., M and L(iter), construct A=XL(iter)XT, B=XD(iter)XTAnd AV=λ BV generalized eigenvalue feature vectors are solved, wherein the feature vector V corresponding to minimum eigenvalue λ is
(4) iter=iter+1, return to step (3) is until meeting stop condition.
(5)WELppIt is equal to finally obtainI.e.
The principal component analytical method comprises the following specific steps that：
To the sample matrix X after standardization*Carry out equalization processing：
Wherein
Obtain equalization treated sample matrixEnable projection after new coordinate be：Then Y ∈ R1 ×M, corresponding variance is：
It is after equalization is handled that wherein E, which takes expectation, ∑,Variance.Specific optimization problem is described as follows：
Pay attention to PPcaIt is vector, utilizes vectorial Rule for derivation： L(PPca)=- PPca T∑PPca+λ(PPca TPPca-1)
After derivation zero setting：
It enablesCalculate characteristic equation：∑PPca=λ PPca, since ∑ is real symmetric matrix, it can be acquired Feature vector corresponding to maximum eigenvalue is：PPca.Embodiment 2：
In order to verify the performance of entropy standardization locality preserving projections analysis, using in UCI data sets in experiment Wine data sets carry out Dimension Reduction Analysis and are compared analysis with tradition LPP algorithms and PCA algorithms.Wine data sets include 3 Classification, number are respectively：59,81,48.Luv space data have 13 attributive character.It is 2 that dimensionality reduction dimension, which is arranged, arest neighbors Number is 1, and thermonuclear parameter is 1.Dimensionality reduction result such as Fig. 1 of algorithms of different wine data sets, shown in 2,3,4.Wherein Fig. 1,2,3, 4 be respectively tradition LPP algorithms, entropy standardization locality preserving projections algorithm iteration 1 time, entropy standardization locality preserving projections algorithm The dimensionality reduction result of iteration 300 times and PCA algorithms.From the comparison of test result we it can be found that tradition LPP algorithms at it In occur juxtaposition phenomenon between two classifications, and the result after entropy standardization locality preserving projections algorithm iteration 300 times The result obtained compared with entropy standardization locality preserving projections algorithm iteration 1 time and tradition LPP algorithms all improves significantly, two classes No longer there is juxtaposition phenomenon between not, this shows that the local neighbor holding structure of entropy standardization locality preserving projections algorithm is excellent In traditional LPP algorithms.In addition, can be seen that PCA only considers the global structure feature of holding initial data from the result of Fig. 4, make There is the juxtaposition phenomenon between multiple and different categorical datas in the data distribution obtained after must projecting.The test result demonstrates PCA is that a global structure keeps algorithm, does not have local neighbours' structure holding capacity.
Fig. 5 shows the situation of change of the target function value of entropy standardization locality preserving projections algorithm iteration 300 times, is not difficult It was found that the entropy standardization locality preserving projections algorithm solution of the present invention belongs to Optimal gradient descent algorithm.Wherein target function value For：
In order to verify influence of the initial parameter to traditional LPP algorithm performances, We conducted arest neighbors be k=1,5,10, Partial structurtes holding capacity contrast test under the setting of 15 different parameters, and with the result of entropy standardization locality preserving projections algorithm It is compared.Test result such as Fig. 6,7,8,9 and 10, shown in 11,12,13.Wherein Fig. 6,7,8,9 correspond to for tradition LPP algorithms Projection coefficient distribution results after the dimensionality reduction of different arest neighbors numbers, Figure 10,11,12,13 calculate for entropy standardization locality preserving projections Method corresponds to projection coefficient distribution results after the dimensionality reductions of different arest neighbors numbers.From the test result of traditional LPP algorithms it is not difficult to find that Fig. 8 and Fig. 9 works as k=10, and result when 15 is better than Fig. 6 and Fig. 7 and works as k=1, when 5 as a result, the test result shows to pass The local neighborhood holding capacity of system LPP algorithms is affected by parameter k value, and in comparison, entropy standardization part keeps throwing Test result of the shadow algorithm under different value of K does not have significant change and local neighborhood holding capacity is better than tradition LPP algorithms, this says Bright entropy standardization locality preserving projections algorithm efficiently solves the problems, such as traditional LPP algorithms to parameter sensitivity and with preferable Robustness.
Embodiment 3：
It is based on entropy standardization locality preserving projections city logistics level evaluation method in order to verify, this experiment is to Shandong Province master The logistics level in city is wanted to be evaluated.Evaluation index is all from Shandong Province's statistical yearbook, has respectively：(1) urban economy development It is horizontal：1. GDP (hundred million yuan)；2. GDP per capita (member)；3. GDP growth rate (%)；4. total industrial output value (hundred million yuan).(2) information-based Degree：1. postal service total amount (hundred million yuan).(3) urban consumption is horizontal：1. the total retail sales of consumer goods (hundred million yuan)；2. wholesale Total value (hundred million yuan) is bought with retail business commodity.(4) Developing Logistics are horizontal：1. communications and transportation, storage and postal industry practitioner (ten thousand People)；2. wholesale and retail business commodity are bought, sell, inventory's number of employees；3. the volume of goods transported (ten thousand tons)；4. (million tons every for volume of the circular flow Kilometer)；5. lorry quantity () for money；6. highway mileage (kilometer)；7. highway mileage (kilometer).Select Shandong Province Case of representational 17 cities as city logistics assessment of levels, specific index value are as shown in figure 14.Most through maximum Small standardization treated city logistics assessment of levels index value X*As shown in figure 15.In order to intuitively illustrate the effect of sequence, Artificially defined two virtual cities：Optimum virtual city and worst virtual city.
Due to dimension difference, each index cannot be directly added；There are certain correlations between each index simultaneously, are directly added The overlapping of information can be increased；In addition it needs to consider respective weight when each index is added, otherwise can not directly carry out synthesis and comment Valence.There is certain correlation in order to illustrate each index, use Charles Bartlett sphericity test method to carry out autocorrelation inspection here It tests.Assume initially that the auto-correlation coefficient matrix of achievement data is unit battle array, the wherein level of signifiance is set as 0.05, P=0<0.05, As a result refusal is assumed, cognizance index data have certain correlation.Auto-correlation coefficient matrix therein is as shown in figure 16.
It determines between each index and carries out dimensionality reduction first with PCA after there is correlation, dimensionality reduction dimension is 1, acquires feature Value and feature vector, and descending sort is carried out to characteristic value.Select the feature vector corresponding to maximum characteristic value as projection Vectorial PPca, and acquire projection of the sample matrix on character pair vector and obtain X ＇Pca=PPca TX*；Then entropy specification is recycled Change locality preserving projections ELpp and carry out dimensionality reduction, dimensionality reduction dimension is 1, acquires eigen vector, and rise to characteristic value Sequence sorts.Select the feature vector corresponding to minimum characteristic value as projection vector WELpp, and sample matrix is acquired in correspondence Projection in feature vector obtains X 'ELpp=WELpp TX*, the X ' that will acquirePcaWith X 'ELppIt is weighted processing and obtains final comment Valence coefficient vector X ', to being final city logistics assessment of levels result after X ' carry out descending arrangements.Result of calculation such as Figure 17 institutes Show, wherein optimum virtual city and worst virtual city has come first and last position respectively.Simultaneously according to initial data On comparison result and associated specialist evaluation it is found that the result that the city logistics level evaluation method that the present invention uses obtains It being consistent with real result, i.e., it includes Qingdao, Jinan, Linyi, Yantai City that Shandong Province, which has the city of best logistics level, Majority belongs to economically developed city of bordering on the sea.Slightly worse city is：Dongying, sunshine, Zaozhuang, Laiwu City belong to economy and owe The inland city reached.
Embodiment 3：
It is based on entropy standardization locality preserving projections city logistics level evaluation method in order to verify, this experiment is to Jilin Province master The logistics level in city is wanted to be evaluated.Evaluation index is all from Jilin Province's statistical yearbook, has respectively：(1) urban economy development It is horizontal：1. GDP (hundred million yuan)；2. GDP per capita (member)；3. the total retail sales of consumer goods (hundred million yuan)；4. total industrial output value (hundred million yuan). (2) urban consumption is horizontal：1. wholesale and retail business commodity buy total value (hundred million yuan).(4) Developing Logistics are horizontal：1. the volume of goods transported (ten thousand Ton)；4. volume of the circular flow (million tons every kilometer)；Select representational 9 cities in Jilin Province as city logistics assessment of levels Case, specific index value are as shown in figure 18.Through maxmin criterionization treated city logistics assessment of levels index value X* As shown in figure 19.In order to intuitively illustrate the effect of sequence, artificially defined two virtual cities：Optimum virtual city and worst void Quasi- city.
Due to dimension difference, each index cannot be directly added；There are certain correlations between each index simultaneously, are directly added Increase the overlapping of information；In addition it needs to consider respective weight when each index is added, otherwise can not directly carry out synthesis and comment Valence.There is certain correlation in order to illustrate each index, use Charles Bartlett sphericity test method to carry out autocorrelation inspection here It tests, assumes initially that the auto-correlation coefficient matrix of achievement data is unit battle array, the wherein level of signifiance is set as 0.05, P=0<0.05, As a result refusal is assumed, cognizance index data have certain correlation.Auto-correlation coefficient matrix therein is as shown in figure 20.
It determines between each index and carries out dimensionality reduction first with PCA after there is correlation, dimensionality reduction dimension is 1, acquires feature Value and feature vector, and descending sort is carried out to characteristic value.Select the feature vector corresponding to maximum characteristic value as projection Vectorial PPca, and acquire projection of the sample matrix on character pair vector and obtain X 'Pca=PPca TX*；Then entropy specification is recycled Change locality preserving projections ELpp and carry out dimensionality reduction, dimensionality reduction dimension is 1, acquires eigen vector, and rise to characteristic value Sequence sorts.Select the feature vector corresponding to minimum characteristic value as projection vector WELpp, and sample matrix is acquired in correspondence Projection in feature vector obtains X 'ELpp=WELpp TX*, the X ' that will acquirePcaWith X 'ELppIt is weighted processing and obtains final comment Valence coefficient vector X ', to being final city logistics assessment of levels result after X ' carry out descending arrangements.Result of calculation such as Figure 21 institutes Show, wherein optimum virtual city and worst virtual city has come first and last position respectively, while according to initial data On comparison result and associated specialist evaluation it is found that the result that the city logistics level evaluation method that the present invention uses obtains It is consistent with real result, i.e., it includes Changchun, Jilin, Siping City, Songyuan City that Jilin Province, which has the city of best logistics level, more Number belongs to the more flourishing city of economy.Slightly worse city is：Side, Tonghua, Bai Shan, Baicheng are prolonged in Liaoyuan, belong to underdeveloped City.

Claims (6)

1. one kind is based on entropy standardization locality preserving projections city logistics level evaluation method.It is characterized in that：This method includes Following steps：
(1) related city logistics are determined by government statistics yearbook or in the way of with related field expert exchanging interview and questionnaire survey Assessment of levels criterion, if N number of evaluation index altogether；
(2) evaluation index obtained to step (1) is classified according to output index and input-occupancy-output analysis, is commented if sharing L output Valence index and P input evaluation index, wherein N=L+P；
(3) the classification indicators information obtained according to step (2), collects the various index values in relation to city logistics assessment of levels, if There are M city, X ∈ RN×M；
(4) the city logistics assessment of levels index value obtained to step (3) carries out numeralization processing, generates sample matrix, will grind The inverted processing of input-occupancy-output analysis information for studying carefully city, is then standardized each index value, makes the number of each index It is worth range to determine between [0,1]；
(5) the sample matrix X after the horizontal index value standardization of the city logistics for the participation evaluation for obtaining step (4)*∈RN×M, Entropy standardization locality preserving projections analysis and principal component analysis are carried out, the pivot number and part that wherein principal component analysis retains It is respectively 1 to keep the projection vector number of projection.It is final to determine PPcaAnd WELppProjection vector, PPca, WELpp∈RN×1；
(6) the projection vector P obtained in step (5) is utilizedPcaSolve overall situation distribution characteristics coefficient vector X ＇Pca=PPca TX* and Local distribution characteristic coefficient vector X 'ELpp=WELpp TX*, X 'Pca, X 'ELpp∈R1×M；
(7) the global distribution characteristics coefficient vector X ' to being acquired in step (6)PcaDrawn game portion distribution characteristics coefficient vector X 'ELppInto Row weighting handles to obtain final evaluation coefficient vector：X '=X 'Pca+βX′ELpp, X ' ∈ R1×M.It is recommended that X ', drops in β=1 Sequence arranges, as final N number of city logistics assessment of levels result.
2. according to claim 1 based on entropy standardization locality preserving projections city logistics level evaluation method, feature It is, output index refers to that can reflect that city logistics infrastructure situation, development of logistics line are horizontal, information-based in step (2) The index of degree and the level of economic development, export-oriented degree and the level of consumption, value show that more greatly the level of city logistics is stronger. And input-occupancy-output analysis refers to city logistics development negative effect caused by society, logistic industry pollutant emission level, traffic is stifled Fill in situation, vehicle, steamer and aircraft noise level, the vibrations of lorry and the damaged condition of road pavement etc., value it is the smaller the better.
3. according to claim 1 based on entropy standardization locality preserving projections city logistics level evaluation method, feature It is, to the inverted processing of input-occupancy-output analysis in step (4), xij∈ X, i=1,2 ..., P, j=1,2 ..., M, xij=1/xij。 Method is maximin method used by being standardized to each index value, is as follows：If xij∈ X, i =1,2 ..., N, j=1,2 ..., M, to arbitrary index xijMethod is as follows used by being standardized：
4. according to claim 1 based on entropy standardization locality preserving projections city logistics level evaluation method, feature It is, P in step (5)PcaComputational methods it is as described below：
To the sample matrix X after standardization*Carry out equalization processing：
Wherein
Obtain equalization treated sample matrixMaximum global characteristics can be kept in order to ask, that is, meet projection coefficient variance most The projection vector P changed greatlyPca, enableCalculate characteristic equation：∑PPca=λ PPca, since ∑ is real symmetric matrix, because This can acquire the feature vector corresponding to its maximum eigenvalue i.e.：PPca。
5. according to claim 1 based on entropy standardization locality preserving projections city logistics level evaluation method, feature It is, W in step (5)ELppComputational methods are as described below：
According to the sample matrix X after standardization*Construct adjacency matrix U(1)∈RM×MWith metric matrix D(1)∈RM×M, wherein uij (1)∈ U(1)Expression formula it is as follows：
xi=[x1i, x2i..., xNi] T, xj=[x1j, x2j..., xNj]T,
I, j=1,2 ..., M, because the method in the present invention is insensitive to the selection of σ and K, it is therefore proposed that σ=1 or 0.5, are arranged K =5.
Construct metric matrix D(1), dij (1)=0, i ≠ j,And acquire drawing This matrix L of pula(1), wherein L(1)=D(1)-U(1)。
Construct A=XL(1)XT, B=XD(1)XTAnd AV=λ BV generalized eigenvalue feature vectors are solved, wherein minimum eigenvalue λ institute Corresponding feature vector V is
Maximum cycle Maxlter=300 and current iteration number iter=2 is set, as iter≤Maxlter
Reconfigure U(iter), according to U(iter)Construct metric matrix D(iter), dij (iter)=0, i ≠ j,And L(iter), construct A=XL(iter)XT, B=XD(iter)XT And AV=λ BV generalized eigenvalue feature vectors are solved, wherein the feature vector V corresponding to minimum eigenvalue λ isIter=iter+1 is repeated until meeting stop condition.WELppIt is equal to finally obtainI.e.
6. according to claim 1 based on entropy standardization locality preserving projections city logistics level evaluation method, feature It is, the projection vector P that step (5) acquiresPcaThe global distributed architecture in luv space can farthest be retained, that is, projected Maximum variance afterwards is conducive to the performance evaluation comparison in macro-scope；Projection vector W in step (5)ELppIt can be utmostly Ground retains the local neighborhood structure in luv space, that is, the local neighborhood after projecting remains unchanged, is conducive in microscopic ranges Performance evaluation compares.
CN201810376569.6A 2018-04-25 2018-04-25 One kind is based on entropy standardization locality preserving projections city logistics level evaluation method Pending CN108596488A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810376569.6A CN108596488A (en) 2018-04-25 2018-04-25 One kind is based on entropy standardization locality preserving projections city logistics level evaluation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810376569.6A CN108596488A (en) 2018-04-25 2018-04-25 One kind is based on entropy standardization locality preserving projections city logistics level evaluation method

Publications (1)

Publication Number Publication Date
CN108596488A true CN108596488A (en) 2018-09-28

Family

ID=63609659

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810376569.6A Pending CN108596488A (en) 2018-04-25 2018-04-25 One kind is based on entropy standardization locality preserving projections city logistics level evaluation method

Country Status (1)

Country Link
CN (1) CN108596488A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111709073A (en) * 2020-06-04 2020-09-25 黄淮学院 Method and system for measuring natural vibration frequency and vibration mode for building structure dynamic analysis

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111709073A (en) * 2020-06-04 2020-09-25 黄淮学院 Method and system for measuring natural vibration frequency and vibration mode for building structure dynamic analysis

Similar Documents

Publication Publication Date Title
Alduayj et al. Predicting employee attrition using machine learning
Duan et al. Test-sheet composition using analytic hierarchy process and hybrid metaheuristic algorithm TS/BBO
Yeo et al. Clustering technique for risk classification and prediction of claim costs in the automobile insurance industry
Gupta et al. Clustering-Classification based prediction of stock market future prediction
Georgescu Using genetic algorithms to evolve type-2 fuzzy logic systems for predicting bankruptcy
US20220156117A1 (en) Systems and methods of optimizing resource allocation using machine learning and predictive control
CN108596488A (en) One kind is based on entropy standardization locality preserving projections city logistics level evaluation method
Wang et al. Research on hybrid real-time picking routing optimization based on multiple picking stations
Shao-hong et al. Location algorithm of transfer stations based on density peak and outlier detection
Chen et al. Simulation and optimization
Luo et al. Finding top-k solutions for the decision-maker in multiobjective optimization
Lai et al. A new method for stock price prediction based on MRFs and SSVM
Keskin A novel perspective to sustainable supplier performance evaluation problem: A case study based on determining the optimum number of clusters
Sbastian et al. Implementation of multi criteria decision making (MCDM) fuzzy neutrosophic TOPSIS-CRITIC in determining sustainability aspects of the location of IoT based products warehouse
Saladi et al. An Enhanced Bankruptcy Prediction Model Using Fuzzy Clustering Model and Random Forest Algorithm.
Wicaksono Priority Modeling for Public Urban Park Development in Feasible Locations Using GIS, Intuitionistic Fuzzy AHP, and Fuzzy TOPSIS
Choong et al. Jack and Masters of All Trades: One-Pass Learning of a Set of Model Sets from Foundation AI Models
Wei et al. A new dynamic credit scoring model based on clustering ensemble
Milosavljević et al. The selection of optimal transport vehicle using multi criteria decision making methods
Lam et al. Adaptive Data Fusion for Multi-task Non-smooth Optimization
Wang Financial distress prediction for listed enterprises using fuzzy C-means
Shimizu et al. Multi-objective optimization for creating a low-carbon logistics system through community-based action
Chen et al. An efficient and effective approach for mining a group stock portfolio using mapreduce
Cholissodin et al. Optimization Deep Intelligence Investment Framework: A New Economic Design for Enhancing and Diversifying Investment Activities in The Kingdom of Saudi Arabia
Mozafari et al. SEGMENTATION OF WATER SUBSCRIBERS BASED ON THEIR LIFETIME VALUE AND RFM MODEL USING DATA MINING

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20180928