CN111178580A - Supermarket site selection method based on improved BP neural network - Google Patents

Supermarket site selection method based on improved BP neural network Download PDF

Info

Publication number
CN111178580A
CN111178580A CN201911187133.3A CN201911187133A CN111178580A CN 111178580 A CN111178580 A CN 111178580A CN 201911187133 A CN201911187133 A CN 201911187133A CN 111178580 A CN111178580 A CN 111178580A
Authority
CN
China
Prior art keywords
layer
neural network
network
output
evaluation
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
CN201911187133.3A
Other languages
Chinese (zh)
Inventor
张贵军
李亭
陈驰
陈芳
卢升荣
刘俊
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.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
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 Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN201911187133.3A priority Critical patent/CN111178580A/en
Publication of CN111178580A publication Critical patent/CN111178580A/en
Pending legal-status Critical Current

Links

Images

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
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G06Q10/043—Optimisation of two dimensional placement, e.g. cutting of clothes or wood
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Development Economics (AREA)
  • Evolutionary Computation (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Health & Medical Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Educational Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Feedback Control In General (AREA)

Abstract

A supermarket site selection method based on an improved BP neural network comprises the steps of firstly, combining a GIS technology, and obtaining the traffic convenience degree of a point to be selected based on the road network condition of a certain area of a certain city; then, analyzing influence factors of supermarket site selection, establishing a BP neural network model, improving the weight and the threshold of the BP neural network model through a genetic algorithm, and improving the learning speed and the learning capacity of the BP neural network; and finally, training the BP neural network, and obtaining the final address point by passing the point to be selected through the trained neural network model. The invention provides a supermarket address selection method with high addressing efficiency based on an improved BP neural network.

Description

Supermarket site selection method based on improved BP neural network
Technical Field
The invention relates to the fields of geographic information data processing technology, network analysis and computer application, in particular to a supermarket site selection method based on an improved BP neural network.
Background
With the development of economy and the improvement of living standard, the demand of people for good life is continuously improved. People's daily life can not leave clothes and eating the resident, and the necessities of daily life generally rely on retail enterprises, and the supermarket is one of people's daily shopping place more. The supermarket is large in area, multiple in types, proper in price and large in service radius, can meet the one-time purchase demand of consumers, and plays an important role in meeting the living needs of people.
Meanwhile, supermarkets generally have large investment scale and long investment period, and once a store site is selected, the store site is difficult to change, the site selection of the supermarket directly influences the investment benefit, so that the supermarket further develops, and the site selection error directly causes great loss. The business area and range of the supermarket have great influence on the operation effect. Therefore, the site selection of the supermarket needs to be considered, and the site selection is necessary according to a scientific method. The site selection of the supermarket is directly related to strategic decision of supermarket operation, is important embodiment for carrying out the view taking consumers as centers, and is a decisive factor influencing the benefit of the supermarket. Unreasonable site selection may cause low investment efficiency, low economic benefit and waste of social resources. If the supermarket can select the site scientifically and reasonably, the economic benefit of the supermarket can be greatly improved, the business range effect is maximized, and the investment benefit is improved. Factors influencing supermarket site selection are many and complex, such as population quantity, population density, rent expense, competitor quantity and the like. It is necessary to process and analyze these influencing factors according to the requirements.
Disclosure of Invention
In order to overcome the defects that the traditional supermarket site selection method only depends on simple data statistics, investigation and analysis, personal experience and subjective judgment of operators, the invention provides a supermarket site selection method based on an improved BP neural network, a GIS technology is combined, an evaluation model of supermarket site selection is established by improving the BP neural network, the supermarket site selection method based on the improved BP neural network can more scientifically and reasonably guide the supermarket site selection, the addressing efficiency is higher, and the addressing effect is more satisfactory.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a supermarket site selection method based on an improved BP neural network comprises the following steps:
1) data acquisition and processing: collecting road information data of a research area, importing road network data of an area to be addressed into ArcGIS Map software to obtain a road network distribution Map of the area, constructing an evaluation model of road traffic convenience degree C, and setting Q ═ k1,k2,...,kmR is { j } is a set of evaluation indexes C, m is the number of evaluation indexes, and R is1,j2,...,jnThe evaluation index k is a set of evaluation levels of each evaluation index k, and n is the number of the evaluation levels;
2) constructing an evaluation matrix of the road traffic convenience degree C, constructing an evaluation matrix G by an evaluation index set of the C, determining a fuzzy relation G from a set Q to a set R by each evaluation object,
Figure BDA0002292661520000021
wherein r isijExpressing the degree of relation of each evaluation index k in the set Q to each evaluation grade j in the set R, and normalizing the evaluation indexes to meet sigma Rij＝1；
3) Determining the input and the output of the BP neural network: analyzing the site selection influence factors, and selecting 6 influence factors in 5 aspects including demand factors, cost factors, competition factors, infrastructure factors and traffic factors as input variables of the neural network, wherein the 6 influence factors are respectively as follows: the number of population, population density, rent expense, the number of competitors, the number of infrastructure and the convenience degree of traffic, and the influence factors are used as the input of the network, and the quality of the operating condition of the supermarket represents the quality of the site selection of the supermarket, so that the output of the neural network is determined as the operating condition of the supermarket;
4) number of hidden layersAnd determining the number of neurons, and establishing a three-layer BP neural network structure, namely an input layer, a hidden layer and an output layer, by combining application research of the BP neural network in the address selection field, wherein the number of the neurons of the input layer is 6, the number of the neurons of the output layer is 1, the number of the nodes of the hidden layer is related to the accuracy and the learning efficiency of the whole BP neural network, and for the BP neural network of the three-layer network structure, the number of the nodes of the hidden layer
Figure BDA0002292661520000022
Wherein n is the number of input nodes, m is the number of output nodes,
Figure BDA0002292661520000023
taking natural numbers between 1 and 10, wherein the learning rate of the network is α, the target error is β, the training times are lambda, and the final number of hidden layer nodes is determined to be y through network training;
5) optimizing BP network connection weight by adopting a genetic algorithm, wherein the process is as follows:
5.1) chromosome coding and population initialization, determining an initial population scale n, coding a weight of a network by adopting a real number coding mode according to a determined network structure, wherein the weight of the network consists of four parts, namely a connection weight between an input layer and a hidden layer, a threshold of the hidden layer, a connection weight between the hidden layer and an output layer and a threshold of the output layer, and randomly generating an initial population L (X) within a certain value range1,X2,...,XnFor a three-layer neural network, any set of complete network weights Xi＝(v1i,v2i,v3i,v4i,v5i,v6i,w1i,w2i,...,wyi,b1i,b2i,...,byi) Where i is 1, 2.., n, and the length of the chromosome is l 2y +6, v1i,v2i,v3i,v4i,v5i,v6iIs the connection weight, w, of input layer neurons to the hidden layer1i,w2i,...,wyiIs the connection weight, w, of the hidden layer neuron to the output layer3iIs the output layer connection weight, b1i,b2i,...,byiIs a threshold value, X, of each level nodeiThe total number of the chromosomes is n, and n is the size of the initial population;
5.2) determining an objective function and a fitness function, taking the weight with the minimum network error searched in all evolutions as the objective function, and taking the formula as E (X)*)≤E(Xi)genWherein i is 1,2,.. the n, gen is 1,2,.. the termgen is population evolution algebra, the function value of the fitness function is used as the basis of selection operation, the reciprocal of the objective function is used as the fitness function, a larger coefficient M is introduced to ensure that the fitness function value is not too small, and the fitness function formula is
Figure BDA0002292661520000031
Wherein i is 1, 2.. times.n;
5.3) selecting an operator, determining that the individuals meeting the requirement selected from the parent generation group are inherited to the next generation group through a fitness function, randomly selecting 2N chromosomes from the parent generation group, sorting according to the fitness values of the individuals, selecting the best N individuals to enter the next generation group, repeating the steps, selecting N individuals again to enter the next generation group, and finally forming a new generation group containing 2N chromosomes;
5.4) a crossover operator, setting the crossover probability as pc, randomly generating a probability p of [0,1) interval when carrying out crossover, randomly selecting the start and stop positions of a plurality of continuous genes in the parent chromosome if p is less than pc, ensuring that the selected positions of the two chromosomes are the same, exchanging the positions of the two groups of genes, and finally forming a new generation of offspring genes;
5.5) a mutation operator, setting the mutation probability as pm, randomly generating a probability q of a [0,1) interval when carrying out mutation, and randomly selecting a gene position on the chromosome to carry out mutation operation if q is less than or equal to pm to form a new individual;
5.6) judging whether the algorithm is finished, if the individual meets the target function condition, finishing the algorithm, otherwise, repeating 5.1) -5.5) until the training target meets the condition, and after the algorithm is finished, decoding the optimal solution individual in the final group to obtain the optimized network connection weight;
6) carrying out BP neural network training, assigning a coding value in a chromosome to a network as an initial weight threshold value of the BP neural network, carrying out signal forward propagation on the constructed input layer data, passing through a hidden layer from the input layer, and finally reaching an output layer, wherein the input signal acts on an output node through the hidden layer, and generates forward training data characteristics through nonlinear transformation;
7) calculating BP neural network errors, reversely propagating expected output errors based on data characteristics of a BP neural network model, reversely propagating the errors by calculating mean square error between expected output values and actual output values, and reversely propagating the output errors layer by layer to an input layer through a hidden layer to obtain error signals of each layer so as to provide basis for adjustment of each unit;
8) and processing the position needing address selection by using the trained BP neural network model to obtain corresponding output, and establishing the position through output prediction so as to achieve the aim of address selection evaluation.
The invention has the beneficial effects that: the method is combined with a GIS technology, based on the road network traffic condition of a site selection area, the weight and the threshold of a BP neural network are optimized through a genetic algorithm, a supermarket site selection model is established by utilizing the improved BP neural network, and the optimal position of the supermarket is obtained based on a point to be selected in the site selection area.
Drawings
FIG. 1 is a flow chart of a supermarket locating method based on an improved BP neural network.
Fig. 2 is a diagram of location information displayed on the GIS platform after data is imported.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
Referring to fig. 1 and 2, a supermarket locating method based on an improved BP neural network includes the following steps:
1) data acquisition and processing: collecting road information data of a research area, importing road network data of an area to be addressed into ArcGIS Map software to obtain a road network distribution Map of the area, and constructing a road intersectionAn evaluation model of the defecation ability C is set to Q ═ k1,k2,...,kmR is { j } is a set of evaluation indexes C, m is the number of evaluation indexes, and R is1,j2,...,jnThe evaluation index k is a set of evaluation levels of each evaluation index k, and n is the number of the evaluation levels;
2) constructing an evaluation matrix of the road traffic convenience degree C, constructing an evaluation matrix G by an evaluation index set of the C, determining a fuzzy relation G from a set Q to a set R by each evaluation object,
Figure BDA0002292661520000041
wherein r isijExpressing the degree of relation of each evaluation index k in the set Q to each evaluation grade j in the set R, and normalizing the evaluation indexes to meet sigma Rij＝1；
3) Determining the input and the output of the BP neural network: analyzing the site selection influence factors, and selecting 6 influence factors in 5 aspects including demand factors, cost factors, competition factors, infrastructure factors and traffic factors as input variables of the neural network, wherein the 6 influence factors are respectively as follows: the number of population, population density, rent expense, the number of competitors, the number of infrastructure and the convenience degree of traffic, and the influence factors are used as the input of the network, and the quality of the operating condition of the supermarket represents the quality of the site selection of the supermarket, so that the output of the neural network is determined as the operating condition of the supermarket;
4) determining the number of hidden layer layers and the number of neurons, establishing a three-layer BP neural network structure by combining application research of a BP neural network in the field of site selection, namely an input layer, a hidden layer and an output layer, wherein the number of neurons of the input layer is 6, the number of neurons of the output layer is 1, the number of nodes of the hidden layer is related to the accuracy and the learning efficiency of the whole BP neural network, and for the BP neural network of the three-layer network structure, the number of nodes of the hidden layer
Figure BDA0002292661520000051
Where n is the number of input nodes and m is the number of output nodesThe number of the first and second groups is,
Figure BDA0002292661520000052
taking natural numbers between 1 and 10, wherein the learning rate of the network is α, the target error is β, the training times are lambda, and the final number of hidden layer nodes is determined to be y through network training;
5) optimizing BP network connection weight by adopting a genetic algorithm, wherein the process is as follows:
5.1) chromosome coding and population initialization, determining an initial population scale n, coding a weight of a network by adopting a real number coding mode according to a determined network structure, wherein the weight of the network consists of four parts, namely a connection weight between an input layer and a hidden layer, a threshold of the hidden layer, a connection weight between the hidden layer and an output layer and a threshold of the output layer, and randomly generating an initial population L (X) within a certain value range1,X2,...,XnFor a three-layer neural network, any set of complete network weights Xi＝(v1i,v2i,v3i,v4i,v5i,v6i,w1i,w2i,...,wyi,b1i,b2i,...,byi) Where i is 1, 2.., n, and the length of the chromosome is l 2y +6, v1i,v2i,v3i,v4i,v5i,v6iIs the connection weight, w, of input layer neurons to the hidden layer1i,w2i,...,wyiIs the connection weight, w, of the hidden layer neuron to the output layer3iIs the output layer connection weight, b1i,b2i,...,byiIs a threshold value, X, of each level nodeiThe total number of the chromosomes is n, and n is the size of the initial population;
5.2) determining an objective function and a fitness function, taking the weight with the minimum network error searched in all evolutions as the objective function, and taking the formula as E (X)*)≤E(Xi)genWherein i is 1,2,.. the n, gen is 1,2,.. the termgen, termgen is population evolution algebra, the function value of the fitness function is the basis of selection operation, the reciprocal of the objective function is used as the fitness function, and a larger coefficient M is introduced to ensure that the fitness function is the reciprocal of the objective functionThe fitness function value is not too small, and the fitness function formula is
Figure BDA0002292661520000061
Wherein i is 1, 2.. times.n;
5.3) selecting an operator, determining that the individuals meeting the requirement selected from the parent generation group are inherited to the next generation group through a fitness function, randomly selecting 2N chromosomes from the parent generation group, sorting according to the fitness values of the individuals, selecting the best N individuals to enter the next generation group, repeating the steps, selecting N individuals again to enter the next generation group, and finally forming a new generation group containing 2N chromosomes;
5.4) a crossover operator, setting the crossover probability as pc, randomly generating a probability p of [0,1) interval when carrying out crossover, randomly selecting the start and stop positions of a plurality of continuous genes in the parent chromosome if p is less than or equal to pc, ensuring that the selected positions of the two chromosomes are the same, exchanging the positions of the two groups of genes, and finally forming a new generation of offspring genes;
5.5) a mutation operator, setting the mutation probability as pm, randomly generating a probability q of a [0,1) interval when carrying out mutation, and randomly selecting a gene position on the chromosome to carry out mutation operation if q is less than or equal to pm to form a new individual;
5.6) judging whether the algorithm is finished, if the individual meets the target function condition, finishing the algorithm, otherwise, repeating 5.1) -5.5) until the training target meets the condition, and after the algorithm is finished, decoding the optimal solution individual in the final group to obtain the optimized network connection weight;
6) carrying out BP neural network training, assigning a coding value in a chromosome to a network as an initial weight threshold value of the BP neural network, carrying out signal forward propagation on the constructed input layer data, passing through a hidden layer from the input layer, and finally reaching an output layer, wherein the input signal acts on an output node through the hidden layer, and generates forward training data characteristics through nonlinear transformation;
7) calculating BP neural network errors, reversely propagating expected output errors based on data characteristics of a BP neural network model, reversely propagating the errors by calculating mean square error between expected output values and actual output values, and reversely propagating the output errors layer by layer to an input layer through a hidden layer to obtain error signals of each layer so as to provide basis for adjustment of each unit;
8) and processing the position needing address selection by using the trained BP neural network model to obtain corresponding output, and establishing the position through output prediction so as to achieve the aim of address selection evaluation.
The embodiment takes an arch villa area in Hangzhou city as an example, and a supermarket site selection method based on an improved BP neural network comprises the following steps:
1) data acquisition and processing: collecting road information data of a research area, importing road network data of an area to be addressed into ArcGIS Map software to obtain a road network distribution Map of the area, constructing an evaluation model of road traffic convenience degree C, and setting Q ═ k1,k2,...,kmR is { j } is a set of evaluation indexes C, m is the number of evaluation indexes, and R is1,j2,...,jnThe evaluation index k is a set of evaluation levels of each evaluation index k, and n is the number of the evaluation levels;
2) constructing an evaluation matrix of the road traffic convenience degree C, constructing an evaluation matrix G by an evaluation index set of the C, determining a fuzzy relation G from a set Q to a set R by each evaluation object,
Figure BDA0002292661520000071
wherein r isijRepresenting the degree of relationship of each evaluation index k in the set Q to each evaluation level j in the set R,
normalizing the data to satisfy sigma rij＝1；
3) Determining the input and the output of the BP neural network: analyzing the site selection influence factors, and selecting 6 influence factors in 5 aspects including demand factors, cost factors, competition factors, infrastructure factors and traffic factors as input variables of the neural network, wherein the 6 influence factors are respectively as follows: the number of population, population density, rent expense, the number of competitors, the number of infrastructure and the convenience degree of traffic, and the influence factors are used as the input of the network, and the quality of the operating condition of the supermarket represents the quality of the site selection of the supermarket, so that the output of the neural network is determined as the operating condition of the supermarket;
4) determining the number of hidden layer layers and the number of neurons, establishing a three-layer BP neural network structure by combining application research of a BP neural network in the field of site selection, namely an input layer, a hidden layer and an output layer, wherein the number of neurons of the input layer is 6, the number of neurons of the output layer is 1, the number of nodes of the hidden layer is related to the accuracy and the learning efficiency of the whole BP neural network, and for the BP neural network of the three-layer network structure, the number of nodes of the hidden layer
Figure BDA0002292661520000072
Wherein n is the number of input nodes, m is the number of output nodes,
Figure BDA0002292661520000073
taking natural number between 1 and 10, learning rate of network is 0.05, target error is 10-3The training frequency is 1000, and the final number of hidden layer nodes is determined to be 12 through network training;
5) optimizing BP network connection weight by adopting a genetic algorithm, wherein the process is as follows:
5.1) chromosome coding and population initialization, determining an initial population scale n, coding a weight of a network by adopting a real number coding mode according to a determined network structure, wherein the weight of the network consists of four parts, namely a connection weight between an input layer and a hidden layer, a threshold of the hidden layer, a connection weight between the hidden layer and an output layer and a threshold of the output layer, and randomly generating an initial population L (X) within a certain value range1,X2,...,XnFor a three-layer neural network, any set of complete network weights Xi＝(v1i,v2i,v3i,v4i,v5i,v6i,w1i,w2i,...,wyi,b1i,b2i,...,byi) Wherein i 1,2, n, the length of the chromosome is 30, v1i,v2i,v3i,v4i,v5i,v6iIs the connection weight, w, of input layer neurons to the hidden layer1i,w2i,...,wyiIs the connection weight, w, of the hidden layer neuron to the output layer3iIs the output layer connection weight, b1i,b2i,...,byiIs a threshold value, X, of each level nodeiCorresponding to one chromosome, wherein n is 50, the total number of chromosomes is 50, and 50 is the size of the initial population;
5.2) determining an objective function and a fitness function, taking the weight with the minimum network error searched in all evolutions as the objective function, and taking the formula as E (X)*)≤E(Xi)genWherein, i is 1,2,., n, gen is 1, 2., 50, the function value of the fitness function is used as the basis of the selection operation, the reciprocal of the objective function is used as the fitness function, a larger coefficient M is introduced to ensure that the fitness function value is not too small, and the fitness function formula is
Figure BDA0002292661520000081
Wherein i is 1, 2.. times.n;
5.3) selecting an operator, determining that the individuals meeting the requirement selected from the parent generation group are inherited to the next generation group through a fitness function, randomly selecting 40 chromosomes from the parent generation group, sorting according to the fitness values of the individuals, selecting the best 20 individuals to enter the next generation group, repeating the steps, selecting 20 individuals again to enter the next generation group, and finally forming a new generation group which comprises 40 chromosomes;
5.4) a crossover operator, setting the crossover probability to be 0.9, randomly generating a probability p of a [0,1) interval when carrying out crossover, randomly selecting the start and stop positions of a plurality of continuous genes in the parent chromosome if p is less than or equal to 0.9, ensuring that the selected positions of the two chromosomes are the same, exchanging the positions of the two groups of genes, and finally forming a new generation of offspring genes;
5.5) a mutation operator, setting the mutation probability to be 0.01, randomly generating a probability q of a [0,1) interval when carrying out mutation, and randomly selecting a gene position on the chromosome to carry out mutation operation to form a new individual if the q is less than or equal to 0.01;
5.6) judging whether the algorithm is finished, if the individual meets the target function condition, finishing the algorithm, otherwise, repeating 5.1) -5.5) until the training target meets the condition, and after the algorithm is finished, decoding the optimal solution individual in the final group to obtain the optimized network connection weight;
6) carrying out BP neural network training, assigning a coding value in a chromosome to a network as an initial weight threshold value of the BP neural network, carrying out signal forward propagation on the constructed input layer data, passing through a hidden layer from the input layer, and finally reaching an output layer, wherein the input signal acts on an output node through the hidden layer, and generates forward training data characteristics through nonlinear transformation;
7) calculating BP neural network errors, reversely propagating expected output errors based on data characteristics of a BP neural network model, reversely propagating the errors by calculating mean square error between expected output values and actual output values, and reversely propagating the output errors layer by layer to an input layer through a hidden layer to obtain error signals of each layer so as to provide basis for adjustment of each unit;
8) and processing the position needing address selection by using the trained BP neural network model to obtain corresponding output, and establishing the position through output prediction so as to achieve the aim of address selection evaluation.
The foregoing description is of the preferred embodiment of the present invention, and the present invention is not limited to the above-described embodiment, but can be modified in various ways without departing from the basic idea of the invention and without exceeding the essence of the invention.

Claims (1)

1. A supermarket site selection method based on an improved BP neural network is characterized in that: the address selection method comprises the following steps:
1) data acquisition and processing: collecting road information data of a research area, importing road network data of an area to be addressed into ArcGIS Map software to obtain a road network distribution Map of the area, constructing an evaluation model of road traffic convenience degree C, and setting Q ═ k1,k2,...,kmWith the proviso of CA set of evaluation indexes, m is the number of evaluation indexes, and R is { j ═ j1,j2,...,jnThe evaluation index k is a set of evaluation levels of each evaluation index k, and n is the number of the evaluation levels;
2) constructing an evaluation matrix of the road traffic convenience degree C, constructing an evaluation matrix G by an evaluation index set of the C, determining a fuzzy relation G from a set Q to a set R by each evaluation object,
Figure FDA0002292661510000011
wherein r isijExpressing the degree of relation of each evaluation index k in the set Q to each evaluation grade j in the set R, and normalizing the evaluation indexes to meet sigma Rij＝1；
3) Determining the input and the output of the BP neural network: analyzing the site selection influence factors, and selecting 6 influence factors in 5 aspects including demand factors, cost factors, competition factors, infrastructure factors and traffic factors as input variables of the neural network, wherein the 6 influence factors are respectively as follows: the number of population, population density, rent expense, the number of competitors, the number of infrastructure and the convenience degree of traffic, and the influence factors are used as the input of the network, and the quality of the operating condition of the supermarket represents the quality of the site selection of the supermarket, so that the output of the neural network is determined as the operating condition of the supermarket;
4) determining the number of hidden layer layers and the number of neurons, establishing a three-layer BP neural network structure by combining application research of a BP neural network in the field of site selection, namely an input layer, a hidden layer and an output layer, wherein the number of neurons of the input layer is 6, the number of neurons of the output layer is 1, the number of nodes of the hidden layer is related to the accuracy and the learning efficiency of the whole BP neural network, and for the BP neural network of the three-layer network structure, the number of nodes of the hidden layer
Figure FDA0002292661510000012
Wherein n is the number of input nodes, m is the number of output nodes,
Figure FDA0002292661510000013
taking natural numbers between 1 and 10, wherein the learning rate of the network is α, the target error is β, the training times are lambda, and the final number of hidden layer nodes is determined to be y through network training;
5) optimizing BP network connection weight by adopting a genetic algorithm, wherein the process is as follows;
5.1) chromosome coding and population initialization, determining an initial population scale n, coding a weight of a network by adopting a real number coding mode according to a determined network structure, wherein the weight of the network consists of four parts, namely a connection weight between an input layer and a hidden layer, a threshold of the hidden layer, a connection weight between the hidden layer and an output layer and a threshold of the output layer, and randomly generating an initial population L (X) within a certain value range1,X2,...,XnFor a three-layer neural network, any set of complete network weights Xi＝(v1i,v2i,v3i,v4i,v5i,v6i,w1i,w2i,...,wyi,b1i,b2i,...,byi) Where i is 1, 2.., n, and the length of the chromosome is l 2y +6, v1i,v2i,v3i,v4i,v5i,v6iIs the connection weight, w, of input layer neurons to the hidden layer1i,w2i,...,wyiIs the connection weight, w, of the hidden layer neuron to the output layer3iIs the output layer connection weight, b1i,b2i,...,byiIs a threshold value, X, of each level nodeiThe total number of the chromosomes is n, and n is the size of the initial population;
5.2) determining an objective function and a fitness function, taking the weight with the minimum network error searched in all evolutions as the objective function, and taking the formula as E (X)*)≤E(Xi)genWherein i is 1,2,.. the n, gen is 1,2,.. the termgen is population evolution algebra, the function value of the fitness function is used as the basis of selection operation, the reciprocal of the objective function is used as the fitness function, a larger coefficient M is introduced to ensure that the fitness function value is not too small, and the fitness function formula is
Figure FDA0002292661510000021
Wherein i is 1, 2.. times.n;
5.3) selecting an operator, determining that the individuals meeting the requirement selected from the parent generation group are inherited to the next generation group through a fitness function, randomly selecting 2N chromosomes from the parent generation group, sorting according to the fitness values of the individuals, selecting the best N individuals to enter the next generation group, repeating the steps, selecting N individuals again to enter the next generation group, and finally forming a new generation group containing 2N chromosomes;
5.4) a crossover operator, setting the crossover probability as pc, randomly generating a probability p of [0,1) interval when carrying out crossover, randomly selecting the start and stop positions of a plurality of continuous genes in the parent chromosome if p is less than pc, ensuring that the selected positions of the two chromosomes are the same, exchanging the positions of the two groups of genes, and finally forming a new generation of offspring genes;
5.5) a mutation operator, setting the mutation probability as pm, randomly generating a probability q of a [0,1) interval when carrying out mutation, and randomly selecting a gene position on the chromosome to carry out mutation operation if q is less than or equal to pm to form a new individual;
5.6) judging whether the algorithm is finished, if the individual meets the target function condition, finishing the algorithm, otherwise, repeating 5.1) -5.5) until the training target meets the condition, and after the algorithm is finished, decoding the optimal solution individual in the final group to obtain the optimized network connection weight;
6) carrying out BP neural network training, assigning a coding value in a chromosome to a network as an initial weight threshold value of the BP neural network, carrying out signal forward propagation on the constructed input layer data, passing through a hidden layer from the input layer, and finally reaching an output layer, wherein the input signal acts on an output node through the hidden layer, and generates forward training data characteristics through nonlinear transformation;
7) calculating BP neural network errors, reversely propagating expected output errors based on data characteristics of a BP neural network model, reversely propagating the errors by calculating mean square error between expected output values and actual output values, and reversely propagating the output errors layer by layer to an input layer through a hidden layer to obtain error signals of each layer so as to provide basis for adjustment of each unit;
8) and processing the position needing address selection by using the trained BP neural network model to obtain corresponding output, and establishing the position through output prediction so as to achieve the aim of address selection evaluation.
CN201911187133.3A 2019-11-28 2019-11-28 Supermarket site selection method based on improved BP neural network Pending CN111178580A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911187133.3A CN111178580A (en) 2019-11-28 2019-11-28 Supermarket site selection method based on improved BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911187133.3A CN111178580A (en) 2019-11-28 2019-11-28 Supermarket site selection method based on improved BP neural network

Publications (1)

Publication Number Publication Date
CN111178580A true CN111178580A (en) 2020-05-19

Family

ID=70653718

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911187133.3A Pending CN111178580A (en) 2019-11-28 2019-11-28 Supermarket site selection method based on improved BP neural network

Country Status (1)

Country Link
CN (1) CN111178580A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070551A (en) * 2020-09-15 2020-12-11 亿景智联(北京)科技有限公司 Retail outlet site selection algorithm based on regional analysis
CN112541786A (en) * 2020-12-11 2021-03-23 中信银行股份有限公司 Site selection method and device for network points, electronic equipment and storage medium
CN117974221A (en) * 2024-04-01 2024-05-03 国网江西省电力有限公司南昌供电分公司 Electric vehicle charging station location selection method and system based on artificial intelligence

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106503802A (en) * 2016-10-20 2017-03-15 上海电机学院 A kind of method of utilization genetic algorithm optimization BP neural network system
CN108717642A (en) * 2018-03-30 2018-10-30 浙江工业大学 A kind of Supermarket Location method based on GIS
CN109359162A (en) * 2018-08-28 2019-02-19 浙江工业大学 A kind of school's site selecting method based on GIS
KR20190051235A (en) * 2017-11-06 2019-05-15 제주대학교 산학협력단 Site selection scheme for electric vehicle chargers based on genetic algorithm
CN109992873A (en) * 2019-03-27 2019-07-09 北京交通大学 The mounting design method of station oriented identification

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106503802A (en) * 2016-10-20 2017-03-15 上海电机学院 A kind of method of utilization genetic algorithm optimization BP neural network system
KR20190051235A (en) * 2017-11-06 2019-05-15 제주대학교 산학협력단 Site selection scheme for electric vehicle chargers based on genetic algorithm
CN108717642A (en) * 2018-03-30 2018-10-30 浙江工业大学 A kind of Supermarket Location method based on GIS
CN109359162A (en) * 2018-08-28 2019-02-19 浙江工业大学 A kind of school's site selecting method based on GIS
CN109992873A (en) * 2019-03-27 2019-07-09 北京交通大学 The mounting design method of station oriented identification

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘海芹: "基于GIS与人工神经网络的经济型酒店选址研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
潘浩: "基于模型优化的物流配送中心选址免疫优化算法", 《电子设计工程》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070551A (en) * 2020-09-15 2020-12-11 亿景智联(北京)科技有限公司 Retail outlet site selection algorithm based on regional analysis
CN112541786A (en) * 2020-12-11 2021-03-23 中信银行股份有限公司 Site selection method and device for network points, electronic equipment and storage medium
CN117974221A (en) * 2024-04-01 2024-05-03 国网江西省电力有限公司南昌供电分公司 Electric vehicle charging station location selection method and system based on artificial intelligence

Similar Documents

Publication Publication Date Title
Salmeron et al. Learning fuzzy cognitive maps with modified asexual reproduction optimisation algorithm
CN109657883B (en) Multi-source data driving-based bank outlet site selection recommendation method
Sohn et al. Decision tree based on data envelopment analysis for effective technology commercialization
CN111967971B (en) Bank customer data processing method and device
CN111178580A (en) Supermarket site selection method based on improved BP neural network
Ünel et al. Valuations of building plots using the AHP method
CN109118013A (en) A kind of management data prediction technique, readable storage medium storing program for executing and forecasting system neural network based
Amirteimoori et al. On the environmental performance analysis: a combined fuzzy data envelopment analysis and artificial intelligence algorithms
CN107292534A (en) The yardstick competition evaluation method and device of urban power distribution network long term dynamics investment
CN107315775A (en) A kind of index calculating platform and method
CN109886807A (en) Personalized recommendation method based on risk income management on a kind of P2P platform
CN114580762A (en) Hydrological forecast error correction method based on XGboost
CN115018357A (en) Farmer portrait construction method and system for production performance improvement
CN110807543A (en) Investment portfolio optimization method and device based on group decision intelligent search
Ragapriya et al. Machine Learning Based House Price Prediction Using Modified Extreme Boosting
CN111652413B (en) Industrial power load prediction method based on multi-Agent distributed mass data processing
Wang et al. Computer supported data-driven decisions for service personalization: a variable-scale clustering method
CN114841610A (en) Shared bicycle scheduling method, device, equipment and readable storage medium
CN115269572A (en) Subject item establishment decision recommendation method and device based on mass scientific research data
CN110033376A (en) A kind of assets management method and device, a kind of calculating equipment and storage medium
CN110991600B (en) Drought intelligent prediction method integrating distribution estimation algorithm and extreme learning machine
Furze et al. Mathematical methods to quantify and characterise the primary elements of trophic systems
Luo et al. Intelligent online food delivery system: a dynamic model to generate delivery strategy and tip advice
CN112785022A (en) Method and system for excavating electric energy substitution potential
Chetthamrongchai et al. Hotel Capacity Planning Using Queuing Systems and Meta-Heuristic Algorithms.

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200519