CN106447112A - Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine - Google Patents

Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine Download PDF

Info

Publication number
CN106447112A
CN106447112A CN201610872124.8A CN201610872124A CN106447112A CN 106447112 A CN106447112 A CN 106447112A CN 201610872124 A CN201610872124 A CN 201610872124A CN 106447112 A CN106447112 A CN 106447112A
Authority
CN
China
Prior art keywords
data
industry
value
output value
scale
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
CN201610872124.8A
Other languages
Chinese (zh)
Inventor
刘仁义
张丰
杜震洪
陈程
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 ZJU
Original Assignee
Zhejiang University ZJU
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 ZJU filed Critical Zhejiang University ZJU
Priority to CN201610872124.8A priority Critical patent/CN106447112A/en
Publication of CN106447112A publication Critical patent/CN106447112A/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
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Tourism & Hospitality (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Theoretical Computer Science (AREA)
  • Development Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Educational Administration (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a construction land scale prediction method of multiple cities based on a stack limited Boltzmann machine. The method includes the following steps: 1) collecting construction land scale data of the cities and corresponding impact factors; 2) correcting two impact factors, namely a second industrial output value and a third industrial output value; 3) conducting normalization processing on the construction land scale data of multiple cities and corresponding impact factors; 4) setting delaying time (Lag) value between attribute data and the construction land scale data; 5) constructing a construction land-use scale prediction model of the cities by using the limited Boltzmann machine; 6) setting a continuous data-oriented limited Boltzmann machine training method, 7) inputting the impact factors obtained after normalization processing so as to obtain construction land scale data of the cities in corresponding years. The method improves the problem that overfitting easily occurs in continuous data processing, and the model prediction precision is improved.

Description

Many town sites scale forecast prediction based on Boltzmann machine limited by stacking Method
Technical field
The invention belongs to the reallocation of land with utilize field, relate to a kind of many cities constructive land scale prediction prediction side Method.
Background technology
Since reform and opening-up in 1978, Chinese city rate constantly rises, and ends for the end of the year 2014, and Chinese city rate reaches To 54.77%.Urban sprawl inevitably results in the loss in peri-urban high-quality arable land.How under conditions of meeting economic development Avoid excessive arable land loss it is ensured that Sustainable Use of Soil Resources will be emphasis and the difficult point of land use and management.Its In, constructive land scale prediction exactly is then basis and the core of land use planning.
Constructive land scale is affected by factors such as economy, landform, population and policies, shows uncertain, non- The feature of linear and capable and experienced immunity.Conventional regression analysis method based on linear hypothesis cannot accurately express constructive land scale Relation and its many factor of influence between.Simultaneously because construction land all-round statistics development is later, single city is obtainable Historical data amount is less, and many Nonlinear Time Series trend forecasting methods hardly result in and are widely applied.
Therefore, how to set up the constructive land scale prediction forecast model in many cities, and can more accurately predict The constructive land scale in many cities is most important with planning to Land_use change.
Content of the invention
The purpose of the present invention is to overcome the deficiencies in the prior art, proposes a kind of many cities based on Boltzmann machine limited by stacking City's constructive land scale Forecasting Methodology.
The technical solution adopted in the present invention is as follows：
Based on many town sites scale forecast Forecasting Methodology of Boltzmann machine limited by stacking, comprise the steps：
1) collection of the constructive land scale data in many cities and its corresponding factor of influence, wherein factor of influence includes：Year Last permanent resident population, GDP, the primary industry output value, the secondary industry output value, the tertiary industry output value, whole society's fixed assets, outlet are total Volume, general financial revenue, local revenue, budgetary expenditure of local government, savings deposits of urban and rural residents year end balance, town dweller are per capita Disposable income, the annual per-capita net income for rural residents, the volume of goods transported and railway passenger volume；
2) it is directed to the secondary industry output value, two factors of influence of the tertiary industry output value are modified；
3) to the constructive land scale data in many cities, year end permanent resident population's data, GDP data, primary industry output value number According to, secondary industry output value data, tertiary industry output value data, whole society's fixed assets data, total export data, finance is total receives Enter data, local revenue data, budgetary expenditure of local government data, savings deposits of urban and rural residents year end balance data, cities and towns residence People's per capita disposable income data, annual per-capita net income for rural residents data, volume of goods transported data and railway passenger volume data are returned One change is processed；
4) time delay between the data that sets a property and constructive land scale data；
5) the constructive land scale forecast model in many cities is set up using Boltzmann machine limited by 2 layers of stacking；
6) utilize step 3) in process after constructive land scale data and step 4) in setting time delay corresponding to Factor of influence data come to step 5) in model be trained；Setting ground floor RBM learning rate is 0.1, train epochs It is set to 100；Second layer learning rate is 0.5, and train epochs are set to 500；When being trained to RBM, arrange adjacent two When secondary reconstructed error is less than 0.0001, lower learning rate, when learning rate is less than 0.0001, i.e. deconditioning；
7) utilize step 6) in training after model, input normalization after factor of influence, the corresponding time can be accessed Many town sites scale data；
On the basis of such scheme, each step can adopt following preferred embodiment：
Described step 2) in, row index correction is entered to the secondary industry output value and the tertiary industry output value, its correction formula is such as Under：
I in formula (1), j, t represent industry, city and time respectively；Represent t (t=1,2 ...) j-th city The growth amount of i industry；WithRepresent t and j-th city of t-1 the i-th industry correction value respectively, and initial value Condition iseij t-1For j-th city of t-1 i industry relative to state of development；In formula (2)For i industry Middle ground all minimum values of the output value,For the value that the output value equal in i industry is maximum；Eij t-1Represent the i of j city t-1 The ground equal output value value of industry.
Described step 3) in normalized formula be：
xi'=(ymax-ymin)×(x-xmin)/(xmax-xmin)+ymin(3)
Wherein xi' for i industry current variable normalization after value；X is the value before the normalization of i industry current variable；xminGeneration The minimum of a value of table current variable, xmaxRepresent the maximum of current variable；ymaxValue is 1, yminValue is 0.
In described forecast model, the node number of ground floor is 8, and the node number of the second layer is 4.
Described step 6) in the concrete training formula of model be：
Wherein parameter rj～Bernoulli (p) (6)
Wherein, viRepresent the node value of visual layers, hjRepresent the node value of hidden layer；σ is sigmod function；P be with The ratio of hidden node hidden by machine, and value is 0.5；WijFor the weight on the side between visual layers and hidden layer, bjAnd aiIt is respectively Visual layers and the biasing of hidden layer, n is training total step number.
The invention has the beneficial effects as follows：Employ the economic indicator modification method in many cities, establish the construction in many cities Land scale unifies forecast model, effectively improves the little problem of single town site scale data amount.Devise towards The limited Boltzmann machine training method of continuous data, improves model and is susceptible to over-fitting in process continuous data Problem, improves model prediction accuracy.
Brief description
Many town sites scale forecast model based on Boltzmann machine limited by stacking for the Fig. 1；
Specific embodiment
With reference to the accompanying drawings and examples the present invention is further detailed.
Based on many town sites scale forecast Forecasting Methodology of Boltzmann machine limited by stacking, comprise the steps：
1) the present embodiment have collected Hangzhou, Ningbo City, Jiaxing City, Jinhua, Wenzhou City, Lishui City, Zhoushan and platform The state city constructive land scale data of 2008 to 2013, and have collected year in statistical yearbook within 2009 to 2014 years in Zhejiang Province Last permanent resident population's data, GDP data, primary industry output value data, secondary industry output value data, tertiary industry output value data, complete Social fixed assets data, total export data, general financial revenue data, local revenue data, budgetary expenditure of local government number According to, savings deposits of urban and rural residents year end balance data, urban residents' disposable income per capita data, the annual per-capita net income for rural residents Data, volume of goods transported data, railway passenger volume data.
2) it is directed to the secondary industry output value, two factors of influence of the tertiary industry output value are modified, its correction formula is as follows：
I in formula (1), j, t represent industry, city and time respectively；Represent t (t=1,2 ...) j-th city The growth amount of the i-th industry；WithRepresent t and j-th city of t-1 the i-th industry correction value respectively, and just Value condition iseij t-1For j-th city of t-1 i industry relative to state of development；In formula (2)Produce for i The minimum value of the equal output value in industry,For the value that the output value equal in i industry is maximum；Eij t-1Represent j city t-1 I industry ground equal output value value.
3) to the constructive land scale data in many cities, year end permanent resident population's data, GDP data, primary industry output value number According to, secondary industry output value data, tertiary industry output value data, whole society's fixed assets data, total export data, finance is total receives Enter data, local revenue data, budgetary expenditure of local government data, savings deposits of urban and rural residents year end balance data, cities and towns residence People's per capita disposable income data, annual per-capita net income for rural residents data, volume of goods transported data and railway passenger volume data are returned One change is processed；Normalized formula is：
xi'=(ymax-ymin)×(x-xmin)/(xmax-xmin)+ymin(3)
Wherein xi' for i industry current variable normalization after value；X is the value after the normalization of i industry current variable；xminGeneration The minimum of a value of table current variable, xmaxRepresent the maximum of current variable；ymaxValue is 1, yminValue is 0.
4) time delay between the data that sets a property and constructive land scale data.
5) set up the constructive land scale forecast model in many cities (as Fig. 1 institute using Boltzmann machine limited by 2 layers of stacking Show).In forecast model in the present embodiment, the node number of ground floor is 8, and the node number of the second layer is 4.
6) many town sites scale data has successional feature, in order that model is applied to continuous data, Simultaneously in order to improve Expired Drugs, eliminate the part of binaryzation in limited Boltzmann machine training process, and add Dropout method is improving the phenomenon of model over-fitting.Therefore, using step 3) in process after constructive land scale data and Step 4) in setting the factor of influence data corresponding to time delay come to step 5) in model be trained；Setting first Layer RBM learning rate is 0.1, and train epochs are set to 100；Second layer learning rate is 0.5, and train epochs are set to 500；When When RBM is trained, when adjacent reconstructed error twice is set less than 0.0001, lower learning rate, when learning rate is less than When 0.0001, i.e. deconditioning.The concrete training formula of model is：
Wherein parameter rj～Bernoulli (p) (6)
Wherein, viRepresent the node value of visual layers, hjRepresent the node value of hidden layer；σ is sigmod function；P be with The ratio of hidden node hidden by machine, and value is 0.5；WijFor the weight on the side between visual layers and hidden layer, bjAnd aiIt is respectively Visual layers and the biasing of hidden layer, n is training total step number.
7) utilize step 6) in training after model, input normalization after factor of influence, the corresponding time can be accessed Many town sites scale data.
In the present embodiment, said method is with Zhejiang Province's 8 prefecture-level cities construction land of 2008 to 2012 and related genus Property as model training data, using the constructive land scale in 8 cities in 2013 as test data.Study the construction of next year Land scale as dependent variable to be predicted, using each association attributes variable of the current year as independent variable.To reach using this Year and the purpose of previous years all available data prediction next year land used.For verifying effectiveness of the invention, experiment utilizes Support vector machine method method as a comparison, distinct methods precision of prediction result such as table 1.Table 2 is illustrated and is proposed using the present invention Training method and traditional training method contrasted.
Table 1
Table 2
Can draw, the many urban construction based on Boltzmann machine limited by the stacking proposed by the present invention unified prediction of land used Method can be good at the constructive land scale in many cities is predicted, and it is original in process effectively to improve model Training method is susceptible to the problem of over-fitting processing continuous data.

Claims (5)

1. a kind of based on stacking limited by Boltzmann machine many town sites scale forecast Forecasting Methodology it is characterised in that Comprise the steps：
1) collection of the constructive land scale data in many cities and its corresponding factor of influence, wherein factor of influence includes：Year end is often Live population, GDP, the primary industry output value, the secondary industry output value, the tertiary industry output value, whole society's fixed assets, total export, wealth Political affairs total income, local revenue, budgetary expenditure of local government, savings deposits of urban and rural residents year end balance, town dweller can prop up per capita Join income, the annual per-capita net income for rural residents, the volume of goods transported and railway passenger volume；
2) it is directed to the secondary industry output value, two factors of influence of the tertiary industry output value are modified；
3) to the constructive land scale data in many cities, year end permanent resident population's data, GDP data, primary industry output value data, Two industry production value data, tertiary industry output value data, whole society's fixed assets data, total export data, general financial revenue number According to, local revenue data, budgetary expenditure of local government data, savings deposits of urban and rural residents year end balance data, town dweller people All disposable income data, annual per-capita net income for rural residents data, volume of goods transported data and railway passenger volume data are normalized Process；
4) time delay between the data that sets a property and constructive land scale data；
5) the constructive land scale forecast model in many cities is set up using Boltzmann machine limited by 2 layers of stacking；
6) utilize step 3) in process after constructive land scale data and step 4) in setting the shadow corresponding to time delay Ring factor data come to step 5) in model be trained；Setting ground floor RBM learning rate is 0.1, and train epochs are arranged For 100；Second layer learning rate is 0.5, and train epochs are set to 500；When being trained to RBM, setting is adjacent to be weighed twice When structure error is less than 0.0001, lower learning rate, when learning rate is less than 0.0001, i.e. deconditioning；
7) utilize step 6) in training after model, input normalization after factor of influence, the many cities obtaining the corresponding time are built If land scale data.
2. a kind of many town sites scale forecast based on Boltzmann machine limited by stacking according to claim 1 is pre- Survey method is it is characterised in that described step 2) in, row index correction is entered to the secondary industry output value and the tertiary industry output value, its Correction formula is as follows：
X i j t ′ = X i j t - 1 ′ + r i j t ( 1 - e i j t - 1 ) - - - ( 1 )
e i j t - 1 = ( E i j t - 1 - E i min ) / ( E i max - E i min ) - - - ( 2 )
I in formula (1), j, t represent industry, city and time respectively；Represent that is produced from t (t=1,2 ...) j-th city i-th The growth amount of industry；WithRepresent t and j-th city of t-1 the i-th industry correction value respectively, and initial condition Foreij t-1For j-th city of t-1 i industry relative to state of development；In formula (2)For in i industry All minimum values of the output value,For the value that the output value equal in i industry is maximum；Eij t-1Represent the i industry of j city t-1 Ground equal output value value.
3. a kind of many town sites scale forecast Forecasting Methodology according to claim 1 is it is characterised in that described Step 3) in normalized formula be：
xi'=(ymax-ymin)×(x-xmin)/(xmax-xmin)+ymin(3)
Wherein xi' for i industry current variable normalization after value；X is the value after the normalization of i industry current variable；xminRepresent and work as The minimum of a value of front variable, xmaxRepresent the maximum of current variable；ymaxValue is 1, yminValue is 0.
4. a kind of many town sites scale forecast based on Boltzmann machine limited by stacking according to claim 1 is pre- Survey method it is characterised in that in described forecast model the node number of ground floor be 8, the node number of the second layer is 4.
5. a kind of many town sites scale forecast based on Boltzmann machine limited by stacking according to claim 1 is pre- Survey method is it is characterised in that described step 6) in the concrete training formula of model be：
h j = σ ( Σ i = 1 n v i W i j + b j ) - - - ( 4 )
v i = σ ( Σ j = 1 n ( h j * r j ) W i j + a i ) - - - ( 5 )
Wherein parameter rj～Bernoulli (p) (6)
Wherein, viRepresent the node value of visual layers, hjRepresent the node value of hidden layer；σ is sigmod function；P is hidden at random Hide the ratio of hidden node, value is 0.5；WijFor the weight on the side between visual layers and hidden layer, bjAnd aiIt is respectively visual Layer and the biasing of hidden layer, n is training total step number.
CN201610872124.8A 2016-09-30 2016-09-30 Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine Pending CN106447112A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610872124.8A CN106447112A (en) 2016-09-30 2016-09-30 Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610872124.8A CN106447112A (en) 2016-09-30 2016-09-30 Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine

Publications (1)

Publication Number Publication Date
CN106447112A true CN106447112A (en) 2017-02-22

Family

ID=58172751

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610872124.8A Pending CN106447112A (en) 2016-09-30 2016-09-30 Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine

Country Status (1)

Country Link
CN (1) CN106447112A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108510161A (en) * 2018-03-12 2018-09-07 华南理工大学 Scarce resource distribution method based on ANALOGY OF BOLTZMANN DISTRIBUTION and multiple index evaluation
CN108875772A (en) * 2018-03-30 2018-11-23 浙江大学 A kind of failure modes model and method being limited Boltzmann machine and intensified learning based on the sparse Gauss Bernoulli Jacob of stacking

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108510161A (en) * 2018-03-12 2018-09-07 华南理工大学 Scarce resource distribution method based on ANALOGY OF BOLTZMANN DISTRIBUTION and multiple index evaluation
CN108875772A (en) * 2018-03-30 2018-11-23 浙江大学 A kind of failure modes model and method being limited Boltzmann machine and intensified learning based on the sparse Gauss Bernoulli Jacob of stacking
CN108875772B (en) * 2018-03-30 2020-04-14 浙江大学 Fault classification model and method based on stacked sparse Gaussian Bernoulli limited Boltzmann machine and reinforcement learning

Similar Documents

Publication Publication Date Title
US20170308934A1 (en) Management method of power engineering cost
Wang et al. Spatial effects of carbon emission intensity and regional development in China
CN103226741B (en) Public supply mains tube explosion prediction method
CN105931116A (en) Automated credit scoring system and method based on depth learning mechanism
CN106845717A (en) A kind of energy efficiency evaluation method based on multi-model convergence strategy
Zhang et al. Impact of structure on unified efficiency for Chinese service sector—A two-stage analysis
CN110111024A (en) Scientific and technological achievement market value evaluation method based on AHP fuzzy comprehensive evaluation model
CN104537433A (en) Sold electricity quantity prediction method based on inventory capacities and business expansion characteristics
CN106022614A (en) Data mining method of neural network based on nearest neighbor clustering
CN103886220B (en) Water data discretization method for setting weight based on BP network and Gini coefficient
CN103426027A (en) Intelligent normal pool level optimal selection method based on genetic neural network models
CN107491877A (en) A kind of power network construction project Budget Performance method based on fuzzy overall evaluation
CN105808689A (en) Drainage system entity semantic similarity measurement method based on artificial neural network
CN109272179A (en) A kind of solar power generation returns of investment overall evaluation system implementation method
Wu et al. Convergence of green total factor productivity in China’s service industry
CN106447112A (en) Construction land scale prediction method of multiple cities based on stack limited Boltzmann machine
CN108537581A (en) Based on the GMDH energy consumption Time Series Forecasting Methods selectively combined and device
CN107862384A (en) A kind of method for building up of distribution network load disaggregated model
CN103778333B (en) A kind of Industrial agglomeration space layout simulative optimization screening technique
Zhou et al. Data-driven solutions for building environmental impact assessment
Yu Research and prediction of ecological footprint using machine learning: A case study of China
Hui Analysis of International Marketing Strategy Based on Intelligent Mining Algorithm for Big Data
Wang et al. The unemployment rate forecast model basing on bp neural network
Xu Assessing the Impact of Transportation Infrastructure on Rural Residents' Income: Using the Quantile Regression Approach
Alorzuke et al. The Studies on Efficiency Analysis in Applied Economics Using a Three-Stage Data Envelopment Analysis

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into 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: 20170222