CN105205623A - Public bicycle station dispatch area division method based on interval weak coupling degree - Google Patents

Abstract

The invention discloses a public bicycle station dispatch area division method based on interval weak coupling degree. The public bicycle station dispatch area division method comprises the following steps: at first, acquiring spatial distance data and bicycle borrowing and returning data representing the relationship between stations, then normalizing rij when the spatial distance data and bicycle borrowing and returning data are combined, and combining the borrowing and returning relationship as a weight into a distance matrix; fusing wij representing the borrowing and returning relationship between two stations and dij representing the distance between the stations, and calculating the similarity between the two stations; adopting the AP clustering algorithm to solve a matrix Sim to obtain area division with weak coupling degree. Dispatching between areas is up to the coupling degree of the two areas; in order to lower the dispatching cost, the key point to reduce dispatching between areas is to reduce the coupling degree; a coupling degree objective function R is used for measuring the rationality of the whole dispatch division scheme. The public bicycle station dispatch area division method ensures that internal bicycle borrowing and returning balance can be maintained through area dispatch, the dispatching efficiency in the dispatch area is improved, and the workload of dispatching between the areas is reduced.

Description

Based on the public bicycles website dispatcher-controlled territory division methods of interval weak coupling degree

Technical field

The invention belongs to public bicycles system regions in municipal intelligent traffic system, relate to a kind of public bicycles website dispatcher-controlled territory division methods based on interval weak coupling degree, thus ensure the website clustering method of strong coupling in interval weak coupling, district.For ensureing that subregion effect reaches optimum, consider the logical communication link that the distance contact between website is leased with bicycle simultaneously.

Background technology

The expansion of city size and the growth of population make urban traffic pressure increase severely.The problems such as consequent toxic emission, environmental pollution are more and more outstanding.Public bicycles, because having the advantages such as pollution-free, mobility strong, can alleviate urban traffic pressure effectively, reduces CO2 emission, improves urban environment, has become government and has advocated the travel modal with citizen's accreditation.But, city public bicycle leasing system is in the starting stage, a more critical problem of current existence is, the distribution of bicycle website does not carry out layout according to civic trip rule, the selection of site location lacks certain scientific analysis, thus cause certain site some time period there will be without bicycle can borrow, without room can and also and car hauler intensively cannot carry out bicycle scheduling in time.

The current dispatcher-controlled territory reasons such as span is comparatively large because area is large, website similarity is little, between region, Route Scheduling is complicated, cause the problems such as dispatching efficiency is low, scheduling cost is excessive.Therefore, need newly to split scheduling Regional Gravity, mark off more reasonable and concentrated region and dispatch, thus raise the efficiency, reduce costs.Meanwhile, when zoning, based on space length, in conjunction with the trip rule that citizen use bicycle, scientific and reasonable Region dividing should be carried out.

At present, less in the research in this field both at home and abroad, most research is the single website layout of logistics field, although with the group clustering difference to some extent of multiple bicycle website, the research of some division methods still has certain reference.In these researchs, the main method of proposition comprises: α-cost model, minimum cost model, is maximumly used a model, cum rights k road clustering algorithm, numeric type data and classification type data mixing clustering algorithm etc.It is more concentrated that the research object of these methods all belongs to individual function, the chain that number ratio is less or logistics warehouse, need the relatively intensive and public bicycles website of contact more complicated between individuality for layout, these methods are not be well suited for.

Summary of the invention

The object of the invention is to repartition the dispatcher-controlled territory of city public bicycle website, a kind of public bicycles website dispatcher-controlled territory division methods based on interval weak coupling degree is provided.Make the website similarity of intra-zone maximum, and the interregional degree of coupling reach minimum, thus the resource distribution effectively between equilibrium region, improve public bicycles dispatching efficiency.

The concrete steps of method of the present invention are as follows:

Step (1). establish N{N ₁, N ₂, N ₃n _nit is the set of all public bicycles lease site; M{M ₁, M ₂, M ₃.M _mit is the dispatcher-controlled territory set that will divide; d _ijfor the space length between any two different website i and website j, wherein i, j ∈ n; Computing formula according to Euclidean distance:

$d_{ij}=\sqrt{{(d_{ix}+d_{iy})}^2+{(d_{jx}+d_{jy})}^2}$

Calculate the distance d between any two websites _ij, space length matrix D is is set up to whole bicycle website:

$Dis=\left[\begin{array}{ccccc}{d}_{11}& d_{12}& d_{13}& ...& d_{1n}\\ d_{21}& d_{22}& d_{23}& ...& d_{2n}\\ d_{31}& d_{32}& d_{33}& ...& d_{3n}\\ ...& ...& ...& ...& ...\\ d_{n1}& d_{n1}& d_{n1}& ...& d_{nn}\end{array}\right]$

Step (2). use r _ijrepresent that the bicycle between website i and website j borrow/is also measured, namely slave site i lease give back to website j borrow also amount and slave site j to lease to give back to website i by means of also measuring summation, calculating r _ijvalue then build bicycle by means of going back matrix Rel:

$\mathrm{Re}l=\left[\begin{array}{ccccc}{r}_{11}& r_{12}& r_{13}& ...& r_{1n}\\ r_{21}& r_{22}& r_{23}& ...& r_{2n}\\ r_{31}& r_{32}& r_{33}& ...& r_{3n}\\ ...& ...& ...& ...& ...\\ r_{n1}& r_{n1}& r_{n1}& ...& r_{nn}\end{array}\right]$

Step (3). by space length data with represent the bicycle of relation between website borrow also data in conjunction with time, need r _ijbe normalized, will be attached in distance matrix as weight by means of also relation; Therefore according to following formula by data normalization;

$w_{ij}=1-\frac{r_{ij}-r_{min}}{r_{max}-r_{min}}$

Wherein, r _maxrepresent by means of the greatest member value of going back in matrix Rel, r _minrepresent by means of the least member value of going back in matrix Rel;

D in space length matrix D is _ijlarger expression is more dissimilar, and less expression is more similar; Borrowing between website i and website j also measures r _ijlarger then weight w _ijbe worth less; If w _ijlarger, then illustrate rent between these two websites also relation more infrequently, namely between these two websites, use the citizen of bicycle fewer;

Step (4). will the w borrowing also relation between two websites be represented _ijwith represent site distance from d _ijmerge, calculate the similarity s between two websites _ij, computing formula is:

s _ij＝w _ij×d _ij

S _ijbe worth between larger expression two websites more dissimilar; Calculate the s between all websites _ijvalue obtains matrix S im:

$Sim=\left[\begin{array}{ccccc}{w}_{11}\×d_{11}& w_{12}\×d_{12}& w_{13}\×d_{13}& ...& w_{1n}\×d_{1n}\\ w_{21}\×d_{21}& w_{22}\×d_{22}& w_{23}\×d_{23}& ...& w_{2n}\×d_{2n}\\ w_{31}\×d_{31}& w_{32}\×d_{32}& w_{33}\×d_{33}& ...& w_{3n}\×d_{3n}\\ ...& ...& ...& ...& ...\\ w_{n1}\×d_{n1}& w_{n2}\×d_{n2}& w_{n3}\×d_{n3}& ...& w_{nn}\×d_{nn}\end{array}\right]$

Step (5). use AP clustering algorithm to solve matrix S im, obtain the Region dividing of weak coupling degree;

Step (6). after obtaining preliminary Region dividing result, the size of the degree of coupling between these two regions is depended in the scheduling between region; And in order to save the cost of scheduling, the key reducing interregional scheduling is exactly reduce the degree of coupling; Therefore, the rationality of whole dispatcher-controlled territory splitting scheme is weighed with degree of coupling objective function R:

$R=\frac{\underset{p=1}{\overset{m}{\Σ}}\underset{q=1}{\overset{m}{\Σ}}{Q}_{MpMq}}{T\×m}\×100\%(p\≠q)$

$T=\underset{i=1}{\overset{n}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{r}_{ij}$

Wherein, Q represents that the bicycle between any two dispatcher-controlled territories is borrowed and also measures sum, and T represents that borrowing between all websites also measures summation; M represents the region sum of division; R value represents the size of the average degree of coupling between m region; Result prove use the method can ensure that between zoning, the degree of coupling is minimum, compared with the splitting scheme drawn with additive method, between each region borrow also relation is minimum.

Beneficial effect of the present invention is as follows:

The present invention to hire a car record data according to user, by the space length between website, combining by means of going back relation data between the website using bicycle to produce with user, using AP clustering algorithm to carry out Region dividing to whole city public bicycle website, ensure that between a region, the degree of coupling is minimum.Thus achieve following target:

Effectively prevent in subregion and do not meet actual phenomenon, if a certain region is across rivers and lakes, Mountainous Scenic Areas etc.Each dispatcher-controlled territory can maintain borrowing of inner bicycle and also balance, and car hauler intra-zone dispatching efficiency improves, and interregional despatching work amount reduces.

Accompanying drawing explanation

Also relation schematic diagram is borrowed between Fig. 1 website.

Fig. 2 site distance is from relation schematic diagram.

Concrete embodiment

Below in conjunction with drawings and Examples, the invention will be further described.

Based on the public bicycles website dispatcher-controlled territory division methods of interval weak coupling degree, specifically comprise the steps:

The user stored in public bicycles system (publicbicyclesystem, PBS) record data of hiring a car are cleaned, excavate two kinds of general datas, space length data with by means of going back relation data, as Fig. 1, specific as follows shown in 2:

Step (1). establish N{N ₁, N ₂, N ₃n _nit is the set of all public bicycles lease site; M{M ₁, M ₂, M ₃.M _mit is the dispatcher-controlled territory set that will divide; d _ijfor the space length between any two different website i and website j, wherein i, j ∈ n.Computing formula according to Euclidean distance:

$d_{ij}=\sqrt{{(d_{ix}+d_{iy})}^2+{(d_{jx}+d_{jy})}^2}$

Calculate the distance d between any two websites _ij, space length matrix D is is set up to whole bicycle website:

$Dis=\left[\begin{array}{ccccc}{d}_{11}& d_{12}& d_{13}& ...& d_{1n}\\ d_{21}& d_{22}& d_{23}& ...& d_{2n}\\ d_{31}& d_{32}& d_{33}& ...& d_{3n}\\ ...& ...& ...& ...& ...\\ d_{n1}& d_{n1}& d_{n1}& ...& d_{nn}\end{array}\right]$

Step (2) uses r _ijrepresent that the bicycle between website i and website j borrow/is also measured, namely slave site i lease give back to website j borrow also amount and slave site j to lease to give back to website i by means of also measuring summation, calculating r _ijvalue then build bicycle by means of going back matrix Rel:

$\mathrm{Re}l=\left[\begin{array}{ccccc}{r}_{11}& r_{12}& r_{13}& ...& r_{1n}\\ r_{21}& r_{22}& r_{23}& ...& r_{2n}\\ r_{31}& r_{32}& r_{33}& ...& r_{3n}\\ ...& ...& ...& ...& ...\\ r_{n1}& r_{n1}& r_{n1}& ...& r_{nn}\end{array}\right]$

Step (3) by space length data with represent the bicycle of relation between website borrow also data in conjunction with time, need r _ijbe normalized, will be attached in distance matrix as weight by means of also relation.Therefore according to following formula by data normalization.

$w_{ij}=1-\frac{r_{ij}-r_{min}}{r_{max}-r_{min}}$

Wherein, r _maxrepresent by means of the greatest member value of going back in matrix Rel, r _minrepresent by means of the least member value of going back in matrix Rel.

D in space length matrix D is _ijlarger expression is more dissimilar, and less expression is more similar.Borrowing between website i and website j also measures r _ijlarger then weight w _ijbe worth less.If w _ijlarger, then illustrate rent between these two websites also relation more infrequently, namely between these two websites, use the citizen of bicycle fewer.

Step (4) borrows the w of also relation between two websites by representing _ijwith represent site distance from d _ijmerge, calculate the similarity s between two websites _ij, computing formula is:

s _ij＝w _ij×d _ij

S _ijbe worth between larger expression two websites more dissimilar.Calculate the s between all websites _ijvalue obtains matrix S im:

$Sim=\left[\begin{array}{ccccc}{w}_{11}\×d_{11}& w_{12}\×d_{12}& w_{13}\×d_{13}& ...& w_{1n}\×d_{1n}\\ w_{21}\×d_{21}& w_{22}\×d_{22}& w_{23}\×d_{23}& ...& w_{2n}\×d_{2n}\\ w_{31}\×d_{31}& w_{32}\×d_{32}& w_{33}\×d_{33}& ...& w_{3n}\×d_{3n}\\ ...& ...& ...& ...& ...\\ w_{n1}\×d_{n1}& w_{n2}\×d_{n2}& w_{n3}\×d_{n3}& ...& w_{nn}\×d_{nn}\end{array}\right]$

Step (5) uses AffinityPropagation (AP, attractor propagation algorithm) clustering algorithm to solve matrix S im, obtains the Region dividing of weak coupling degree.Transmitted by the message of two types between each data point in AP algorithm: Responsibility and Availability.Res (i, k) represents the numerical value message being sent to candidate cluster center k from an i, and this value passes judgment on the foundation whether k point is suitable as the cluster centre of i point.Ava (i, k) represents the numerical value message being sent to i from candidate cluster center k, and whether this value passes judgment on i point to select k point as the foundation of its cluster centre.Therefore, res (i, k) and ava (i, k) two value larger, then k point is larger as the possibility of cluster centre, and meanwhile, some i is also under the jurisdiction of class using k as cluster centre with regard to having larger possibility.AP algorithm is used to obtain the area division scheme that degree of coupling R value can be made to reach minimum, website reasonable quantity in each region divided, region area and website quantity relative equilibrium, can ensure that change work is basically identical, and the station associate between region also can be lower, thus ensure interregional weak coupling degree.In addition, independently, can't there is the problem in cross a river stream or mountain region in Dou Shi space, each region; Specific works flow process is as follows:

Step1: calculate the similarity s (i, k) between N number of data point, structure similarity matrix S, because this similarity adopts Euclidean distance as metering, therefore numerical value is got negative, and more namely similarity is larger for value.

Step2: the value of initialized reference degree, given iterations N.

Step3: the value calculating res:

res(i,k)＝s(i,k)-max{ava(i,k')+s(i,k')},k≠k'

Step4: the value calculating ava:

$ava(i,k)=min\{0,res(k,k)+\mathrm{\Σ}max\{0,res(i^{\′},k)\left\}\right\},i^{\′}\∉\{i,k\}$

ava(k,k)＝∑max{0,res(i',k)},i'≠k

Step5: determine that whether this point is as cluster centre according to the value of res (k, k)+ava (k, k).

Step6: stop when cluster centre N continuous time does not change calculating.

After step (6) obtains preliminary Region dividing result, the size of the degree of coupling between these two regions is depended in the scheduling between region.And in order to save the cost of scheduling, the key reducing interregional scheduling is exactly reduce the degree of coupling.Therefore, the rationality of whole dispatcher-controlled territory splitting scheme is weighed with degree of coupling objective function R:

$R=\frac{\underset{p=1}{\overset{m}{\Σ}}\underset{q=1}{\overset{m}{\Σ}}{Q}_{MpMq}}{T\×m}\×100\%(p\≠q)$

$T=\underset{i=1}{\overset{n}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{r}_{ij}$

Wherein, Q represents that the bicycle between any two dispatcher-controlled territories is borrowed and also measures sum, and T represents that borrowing between all websites also measures summation.M represents the region sum of division.R value represents the size of the average degree of coupling between m region.Result prove use the method can ensure that between zoning, the degree of coupling is minimum, compared with the splitting scheme drawn with additive method, between each region borrow also relation is minimum.

Claims (1)

1., based on the public bicycles website dispatcher-controlled territory division methods of interval weak coupling degree, it is characterized in that comprising the steps:

Step (1). establish N{N ₁, N ₂, N ₃n _nit is the set of all public bicycles lease site; M{M ₁, M ₂, M ₃.M _mit is the dispatcher-controlled territory set that will divide; d _ijfor the space length between any two different website i and website j, wherein i, j ∈ n; Computing formula according to Euclidean distance:

$d_{ij}=\sqrt{{(d_{ix}+d_{iy})}^2+{(d_{jx}+d_{jy})}^2}$

Calculate the distance d between any two websites _ij, space length matrix D is is set up to whole bicycle website:

$Dis=\left[\begin{array}{ccccc}{d}_{11}& d_{12}& d_{13}& ...& d_{1n}\\ d_{21}& d_{22}& d_{23}& ...& d_{2n}\\ d_{31}& d_{32}& d_{33}& ...& d_{3n}\\ ...& ...& ...& ...& ...\\ d_{n1}& d_{n1}& d_{n1}& ...& d_{nn}\end{array}\right]$

Step (2). use r _ijrepresent that the bicycle between website i and website j borrow/is also measured, namely slave site i lease give back to website j borrow also amount and slave site j to lease to give back to website i by means of also measuring summation, calculating r _ijvalue then build bicycle by means of going back matrix Rel:

$\mathrm{Re}l=\left[\begin{array}{ccccc}{r}_{11}& r_{12}& r_{13}& ...& r_{1n}\\ r_{21}& r_{22}& r_{23}& ...& r_{2n}\\ r_{31}& r_{32}& r_{33}& ...& r_{3n}\\ ...& ...& ...& ...& ...\\ r_{n1}& r_{n1}& r_{n1}& ...& r_{nn}\end{array}\right]$

Step (3). by space length data with represent the bicycle of relation between website borrow also data in conjunction with time, need r _ijbe normalized, will be attached in distance matrix as weight by means of also relation; Therefore according to following formula by data normalization;

$w_{ij}=1-\frac{r_{ij}-r_{min}}{r_{max}-r_{min}}$

Wherein, r _maxrepresent by means of the greatest member value of going back in matrix Rel, r _minrepresent by means of the least member value of going back in matrix Rel;

D in space length matrix D is _ijlarger expression is more dissimilar, and less expression is more similar; Borrowing between website i and website j also measures r _ijlarger then weight w _ijbe worth less; If w _ijlarger, then illustrate rent between these two websites also relation more infrequently, namely between these two websites, use the citizen of bicycle fewer;

Step (4). will the w borrowing also relation between two websites be represented _ijwith represent site distance from d _ijmerge, calculate the similarity s between two websites _ij, computing formula is:

s _ij＝w _ij×d _ij

S _ijbe worth between larger expression two websites more dissimilar; Calculate the s between all websites _ijvalue obtains matrix S im:

$Sim=\left[\begin{array}{ccccc}{w}_{11}\×d_{11}& w_{12}\×d_{12}& w_{13}\×d_{13}& ...& w_{1n}\×d_{1n}\\ w_{21}\×d_{21}& w_{22}\×d_{22}& w_{23}\×d_{23}& ...& w_{2n}\×d_{2n}\\ w_{31}\×d_{31}& w_{32}\×d_{32}& w_{33}\×d_{33}& ...& w_{3n}\×d_{3n}\\ ...& ...& ...& ...& ...\\ w_{n1}\×d_{n1}& w_{n2}\×d_{n2}& w_{n3}\×d_{n3}& ...& w_{nn}\×d_{nn}\end{array}\right]$

Step (5). use AP clustering algorithm to solve matrix S im, obtain the Region dividing of weak coupling degree;

Step (6). after obtaining preliminary Region dividing result, the size of the degree of coupling between these two regions is depended in the scheduling between region; And in order to save the cost of scheduling, the key reducing interregional scheduling is exactly reduce the degree of coupling; Therefore, the rationality of whole dispatcher-controlled territory splitting scheme is weighed with degree of coupling objective function R:

$R=\frac{\underset{p=1}{\overset{m}{\Σ}}\underset{q=1}{\overset{m}{\Σ}}{Q}_{MpMq}}{T\×m}\×100\%(p\≠q)$

$T=\underset{i=1}{\overset{n}{\Σ}}\underset{j=1}{\overset{n}{\Σ}}{r}_{ij}$

Wherein, Q represents that the bicycle between any two dispatcher-controlled territories is borrowed and also measures sum, and T represents that borrowing between all websites also measures summation; M represents the region sum of division; R value represents the size of the average degree of coupling between m region; Result prove use the method can ensure that between zoning, the degree of coupling is minimum, compared with the splitting scheme drawn with additive method, between each region borrow also relation is minimum.