CN104778508B - A kind of public bicycles based on multisource data fusion lease Forecasting Methodology - Google Patents
A kind of public bicycles based on multisource data fusion lease Forecasting Methodology Download PDFInfo
- Publication number
- CN104778508B CN104778508B CN201510154943.4A CN201510154943A CN104778508B CN 104778508 B CN104778508 B CN 104778508B CN 201510154943 A CN201510154943 A CN 201510154943A CN 104778508 B CN104778508 B CN 104778508B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- msup
- particle
- data
- 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.)
- Expired - Fee Related
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Medicines Containing Antibodies Or Antigens For Use As Internal Diagnostic Agents (AREA)
Abstract
The invention discloses a kind of public bicycles based on multisource data fusion to lease Forecasting Methodology.The present invention first rents public bicycles/and the also historical data such as record, weather, temperature, festivals or holidays passes through data cleansing and pretreatment, acquisition training dataset.Data set is classified using clustering algorithm, the difference for dividing public bicycles leases pattern；Sorted data set is built into Bayes classifier, for leasing pattern according to belonging to following certain day festivals or holidays, weather, the temperature situation prediction same day；Adaptive PSO Neural Network model corresponding to each pattern is respectively trained for the data set of different mode.Finally, certain day is predicted by Bayes classifier to lease PSO Neural Network model corresponding to modal selection and lease rule predict public bicycles.Precision of prediction of the present invention is high, arithmetic speed is fast, can be that user rents offer reference frame of returning the car, reduce the public bicycles website non-equilibrium state duration, improve user satisfaction.
Description
Technical field
The invention belongs to city intelligent public transportation system technical field, is related to a kind of based on the public of multisource data fusion
Bicycle leases Forecasting Methodology, particularly leases what quantity was predicted to different periods bicycle in each public bicycles website
Method.
Background technology
A part of the public bicycles system as urban public transport, has the advantages that pollution-free, mobility strong, can be with
Effectively alleviate urban traffic pressure, reduce CO2 emission, improve urban environment.Due to the mobility and tide of citizens' activities
Nighttide, the problem of no bicycle can be borrowed, can gone back without room occurs in some periods.City ITS
(Intelligent Transportation System, abbreviation ITS) is by advanced information technology, mechanics of communication, sensing skill
Art, control technology and computer technology etc., which are effectively integrated, applies to whole traffic management system, and one set up
Kind in a wide range of, it is comprehensive play a role, in real time, accurately and efficiently multi-transportation and management system.It can be used for solving
The problems that urban public transport field occurs.
Public bicycles lease the part that quantitative forecast is City ITS, each the purpose is to predict exactly
Bicycle leases quantity in public bicycles website different time sections, so as to deploy Operation Measures in advance, effectively solve " rent/also
Car is difficult " problem.
At present, the research both at home and abroad in this field is less, and most of is the prediction for being directed to automotive field.Ground at these
In studying carefully, the main method of proposition includes：History averaging model, autoregression model, multivariate regression models, nonparametric Regression Model,
Kalman filter model, neural network model, supporting vector machine model etc..But due to the particularity of public bicycles, these
Prediction of the method for the field is not to be well suited for.History averaging model and regression analysis model assume that the magnitude of traffic flow is to abide by
Follow strict periodicity, and public bicycles lease that there is the characteristics of uncertain, non-linear, aperiodicity.Kalman filters
Wave pattern is not to be well suited on the traffic flow forecasting of short period of time.Supporting vector machine model is using a large amount of true
Showed during data progress traffic flow forecasting poor.It is a kind of parallel and neural network model is modeled according to input and output
Computation model, arithmetic speed is very fast, and possess good non-linear mapping capability and learn by oneself property, adaptive ability.But
Traditional neural network model there is also convergence rate it is slow, be easily trapped into the shortcomings that local optimum.
The content of the invention
The invention aims to overcome the shortcomings of Individual forecast model, a kind of public affairs based on multisource data fusion are proposed
Bicycle leases Forecasting Methodology altogether.The present invention is improved and merged to conventional method, is used on the basis of cluster analysis
Naive Bayes Classification Algorithm carries out leasing model prediction, the different patterns of leasing is respectively trained respective adaptive population
Neural network model, and the public bicycles leased using these model prediction differences under pattern lease quantity, raising is leased
Prediction accuracy.
The present invention solves comprising the following steps that for the technical scheme that its technical problem uses：
The present invention specifically comprises the following steps：
Step (1) data acquisition and procession：
Read the user that stores in public bicycles system to hire a car record, calculate each website every the remaining public affairs of time period t ime
Bicycle quantity altogether, establishes vectorial Ndate=[n1,n2,...,nj,...,nM], M is period sum；Obtain and go through from internet
History weather conditions, temperature and festivals or holidays situation, establish environment attribute vector S={ S1,S2,...,Si,...,SL, Si=[d, w,
T], L is the number that data sample concentrates sample；M, L, i and j are positive integer, wherein j≤M, i≤L；
Described user hire a car record include lease duration, lease website, lease knee number, the time of returning the car, website of returning the car,
Return the car knee number；
njRepresent remaining bicycle quantity in j-th of time period t ime website；
SiRepresent the environment attribute of i-th day；
D represents the festivals or holidays attribute on the same day, and d value is
W represents weather conditions, and w value is
T represents temperature, and t value is
Step (2) K-means is clustered：
Ndate=[n1,n2,...,nj,...,nM] data tuple being concentrated as data sample to be sorted；From number
According to k point is randomly selected in sample set as initial classes center, μ is used respectively1,μ2,...,μkRepresent class c1,c2,...,ckClass
Center；Wherein, k is positive integer；
K-means clustering algorithms comprise the following steps：
2-1. calculates remaining data tuple NiWith class center μkDistance, by remaining data tuple NiAssign to its away from
From in most short class；
2-2. recalculates all kinds of new class center μk *；
2-3. repeat steps 2-1 and step 2-2, until each data tuple generic all no longer changes；
Step (3) builds Bayes classifier：After being classified using K-means clustering algorithms to set of data samples, obtain
To k class, each data tuple that data sample is concentrated will obtain a class label；Bayes Method is determined based on Bayes
Reason：
Calculate P (X), P (H) and P (X | H)；
The adaptive PSO Neural Network model training of step (4)
Use sorted data sample training affiliated adaptive PSO Neural Network model of all categories, setting nerve
The input layer unit number of network is n, and it is p to imply layer unit number, and output layer unit number is q；Then needed in neutral net
Weights number N=n × p+p × q of adjustment, with the dimension in the weights number N problem of representation space for needing to adjust；With vectorial Xi=
[xi1,xi2,...,xiN] represent all weights of neutral net a possible value, also illustrate that particle in particle swarm optimization algorithm
Position；PSO Neural Network model training step is as follows：
The position of each particle and speed in 4-1. initialization population populations, the position of particle and speed be located at [0,1] it
Between, i.e., G vectorial X of generation at randomi=[xi1,xi2,...,xiN], positions of the 1≤i≤G as particle, xiBetween [0,1]；
Generate G vectorial V at random againi=[vi1,vi2,...,viN], speed of the G+1≤i≤2 × G as particle, viBetween [0,1] it
Between；Set maximum iteration iter max, current iteration number k=1；
4-2. calculates the fitness value of each particle, and it is itself optimal value P to store the current position of each particlebi, store all
The minimum position of fitness value is global optimum P in particlebg；
4-3. judges whether k reaches maximum iteration iter max, if so, going to step 4-7, otherwise goes to 4-4；
4-4. updates each particle rapidity and position, calculates the new fitness value of each particle；If the current adaptation of i-th of particle
Angle value is better than Pbi, then P is updatedbi；If the optimal location of current particle group is better than Pbg, then P is updatedbg；K=k+1 is set；
4-5. updates inertia weight ω according to inertia weight calculation formula；
If 4-6. global optimums position PbgDo not change in 10 iteration, then go to step 4-7, otherwise go to step 4-
4；
4-7. exports PbgWeights as the PSO Neural Network trained；
Step (5) leases model prediction：According to the environment attribute vector S to be predicted of inputpreWith the P (H) being calculated, P
The value of (X | H), calculate the probability that the data tuple belongs to of all categories, the maximum classification of output posterior probability；
Step (6) public bicycles lease prediction
The classification exported in model prediction is leased according to step (5), selects corresponding adaptive PSO Neural Network mould
Type, the remaining public bicycles quantity of preceding n period is inputted into the model, output is the public bicycles quantity of prediction.
Calculation formula is as follows described in step (2)：
N in 2-1iWith class center μkDistance calculation formula it is as follows：
New class center calculation formula is as follows in 2-2：
μk *For class ckNew class center；
|nk| it is class ckThe number of middle data tuple；
NkiFor class ckIn i-th of data tuple.
The calculation formula and variable implication used in step (3) is as follows：
|ck,T| it is to belong to class c in set of data samples TkSample number；
L is the total sample number in set of data samples T；
P (X | H)=P (di|ck)P(wi|ck)P(ti| ck),
For class ckMiddle attribute d value, it is diNumber of samples；
For class ckMiddle attribute w value, it is wiNumber of samples；
For class ckMiddle attribute t value, it is tiNumber of samples.
The calculation formula and variable implication used in step (4) is as follows：
Fitness value fitness (Xg) calculation formula be：
M is that data sample concentrates data tuple number；
xliFor i-th of input of l-th of data sample；
θhjFor the threshold value of j-th of unit of hidden layer；
θokFor the threshold value of k-th of unit of output layer；
wijFor the connection weight of i-th of input layer and j-th of hidden layer；
wjkFor the connection weight of j-th of hidden layer and k-th of output layer；
dlkFor k-th of output of l-th of data sample；
Speed calculation formula is：
Vi kFor particle XiIn the speed of kth time iteration；
r1,r2It is random number of the value between [0,1]；
c1, c2For one group of random number, c is typically taken1=c2=2.5；
ω is inertia weight；
Inertia weight ωkCalculation formula is：
ω max=0.9, ω min=0.2；
The position calculation formula of particle is：
Xi k+1=Xi k+Vi k+1；
Xi kThe position of i-th of particle during iteration secondary for kth.
The calculation formula used in step (6) is as follows：
Neural network model output O is specific as follows：
The present invention has the beneficial effect that：
The present invention hires a car the data such as record, weather, temperature, festivals or holidays according to user, to each website public bicycles flow
It is predicted, rather than just usage history loan data.Present invention generally provides a kind of prediction of multisource data fusion
Method, it can accurately and efficiently predict each website public bicycles volume residual of certain day certain period.Specifically, realize
Following target：
The record that can be hired a car from original user extracts each period each website residue bicycle quantity, is easy to later stage cluster point
Analysis, construction Bayes classifier and training PSO Neural Network.
Multiple data sources can be combined, public bicycles are leased with pattern and carries out static prediction, is advantageous to public bicycles
Operator carries out scheduling preparation in advance.
Dynamic prediction can be carried out according to the remaining bicycle quantity of n period before certain day, improve public bicycles and rent
By means of quantitative forecast accuracy, scheduling time and specific scheduling quantity suggestion are provided for public bicycles operator.It can improve
Promptness is dispatched, reduces scheduling cost.
It is fast that specific implementation result shows as (1) speed of service, completes Bayes classifier in structure and training is adaptive
After PSO Neural Network model, static and dynamic prediction can be carried out within the extremely short time；(2) precision of prediction is high, makes
Simulation and forecast experiment is carried out with True Data, consensus forecast accuracy is more than 96%.
Brief description of the drawings
Fig. 1 public bicycles lease Forecasting Methodology figure.
The adaptive PSO Neural Network algorithm flow charts of Fig. 2.
Specific embodiment
The invention will be further described with reference to the accompanying drawings and examples.
As shown in figure 1, a kind of public bicycles based on multisource data fusion lease Forecasting Methodology, specifically include following several
Individual step：
Step (1) data acquisition and procession：
The user for reading storage in public bicycles system (public bicycle system, PBS) hires a car record, counts
Each website is calculated every time period t ime residue public bicycles quantity, establishes vectorial Ndate=[n1,n2,...,nj,...,nM], M
For period sum；Weather history situation temperature and festivals or holidays situation are obtained from internet, establishes environment attribute vector S=
{S1,S2,...,Si,...,SL, Si=[d, w, t], L are the number that data sample concentrates sample；M, L, i and j are positive integer,
Wherein j≤M, i≤L；
Described user hire a car record include lease duration, lease website, lease knee number, the time of returning the car, website of returning the car,
Return the car knee number；
njRepresent remaining bicycle quantity in j-th of time period t ime website；
SiRepresent the environment attribute of i-th day；
D represents the festivals or holidays attribute on the same day, and d value is
W represents weather conditions, and w value is
T represents temperature, and t value is
Step (2) K-means is clustered：
Ndate=[n1,n2,...,nj,...,nM] data tuple being concentrated as data sample to be sorted；From number
According to k point is randomly selected in sample set as initial classes center, μ is used respectively1,μ2,...,μkRepresent class c1,c2,...,ckClass
Center；Wherein, k is positive integer；
K-means clustering algorithms comprise the following steps：
2-1. calculates remaining data tuple NiWith class center μkDistance, by remaining data tuple NiAssign to its away from
From in most short class；
2-2. recalculates all kinds of new class center μk *；
2-3. repeat steps 2-1 and step 2-2, until each data tuple generic all no longer changes.
Step (3) builds Bayes classifier：After being classified using K-means clustering algorithms to set of data samples, obtain
To k class, each data tuple that data sample is concentrated will obtain a class label.Bayes Method is determined based on Bayes
Reason：
Calculate P (X), P (H) and P (X | H).
Step (4) is as shown in Fig. 2 adaptive PSO Neural Network model training：Instructed using sorted data sample
Practice affiliated adaptive PSO Neural Network model of all categories, set the input layer unit number of neutral net as n, hidden layer
Unit number is p, and output layer unit number is q.The weights number N=n × p+p × q adjusted is then needed in neutral net, with need
The dimension in the weights number N problem of representation space to be adjusted.With vectorial Xi=[xi1,xi2,...,xiN] represent that neutral net owns
One possible value of weights, also illustrate that the position of particle in particle swarm optimization algorithm.PSO Neural Network model training walks
It is rapid as follows：
The position of each particle and speed in 4-1. initialization population populations, the position of particle and speed be located at [0,1] it
Between, i.e., G vectorial X of generation at randomi=[xi1,xi2,...,xiN], positions of the 1≤i≤G as particle, xiBetween [0,1]；
Generate G vectorial V at random againi=[vi1,vi2,...,viN], speed of the G+1≤i≤2 × G as particle, viBetween [0,1] it
Between.Set maximum iteration itermax, current iteration number k=1；
4-2. calculates the fitness value of each particle, and it is itself optimal value P to store the current position of each particlebi, store all
The minimum position of fitness value is global optimum P in particlebg；
4-3. judges whether k reaches maximum iteration itermax, if so, going to step 4-7, otherwise goes to 4-4.
4-4. updates each particle rapidity and position, calculates the new fitness value of each particle.If the current adaptation of i-th of particle
Angle value is better than Pbi, then P is updatedbi.If the optimal location of current particle group is better than Pbg, then P is updatedbg；K=k+1 is set；
4-5. updates inertia weight ω according to inertia weight calculation formula；
If 4-6. global optimums position PbgDo not change in 10 iteration, then go to step 4-7, otherwise go to step 4-
4；
4-7. exports PbgWeights as the PSO Neural Network trained.
Step (5) leases model prediction：According to the environment attribute vector S to be predicted of inputpreWith the P (H) being calculated, P
The value of (X | H), calculate the probability that the data tuple belongs to of all categories, the maximum classification of output posterior probability.
Step (6) public bicycles lease prediction：The classification exported in model prediction is leased according to step (5), selects phase
The adaptive PSO Neural Network model answered, the remaining public bicycles quantity of preceding n period is inputted into the model, exported
The public bicycles quantity as predicted：
Calculation formula is as follows described in step (2)：
N in 2-1iWith class center μkDistance calculation formula it is as follows：
New class center calculation formula is as follows in 2-2：
μk *For class ckNew class center；
|nk| it is class ckThe number of middle data tuple；
NkiFor class ckIn i-th of data tuple.
The calculation formula and variable implication used in step (3) is as follows：
|ck,T| it is to belong to class c in set of data samples TkSample number；
L is the total sample number in set of data samples T.
P (X | H)=P (di|ck)P(wi|ck)P(ti|ck),
For class ckMiddle attribute d value, it is diNumber of samples；
For class ckMiddle attribute w value, it is wiNumber of samples；
For class ckMiddle attribute t value, it is tiNumber of samples；
The calculation formula and variable implication used in step (4) is as follows：
Fitness value fitness (Xg) calculation formula be：
M is that data sample concentrates data tuple number；
xliFor i-th of input of l-th of data sample；
θhjFor the threshold value of j-th of unit of hidden layer；
θokFor the threshold value of k-th of unit of output layer；
wijFor the connection weight of i-th of input layer and j-th of hidden layer；
wjkFor the connection weight of j-th of hidden layer and k-th of output layer；
dlkFor k-th of output of l-th of data sample.
Speed calculation formula is：
Vi kFor particle XiIn the speed of kth time iteration；
r1,r2It is random number of the value between [0,1]；
c1, c2For one group of random number, c is typically taken1=c2=2.5；
ω is inertia weight；
Inertia weight ωkCalculation formula is：
ω max=0.9, ω min=0.2；
The position calculation formula of particle is：
Xi k+1=Xi k+Vi k+1；
Xi kThe position of i-th of particle during iteration secondary for kth；
The calculation formula used in step (6) is as follows：
Neural network model output O is specific as follows：
Claims (4)
1. a kind of public bicycles based on multisource data fusion lease Forecasting Methodology, it is characterised in that comprise the following steps：
Step (1) data acquisition and procession：
Read the user that stores in public bicycles system to hire a car record, it is public every time period t ime residues to calculate each website
Bicycle quantity, vectorial N is established for each websitedate=[n1,n2,...,nj,...,nM]；History day is obtained from internet
Vaporous condition, temperature and festivals or holidays situation, establish environment attribute vector S={ S1,S2,...,Si,...,SL, Si=[d, w, t], L
The number of sample is concentrated for data sample；M, L, i and j are positive integer, wherein j≤M, i≤L；
Described user record of hiring a car includes lease duration, leases website, lease knee number, time of returning the car, website of returning the car, return the car
Knee number；
njRepresent remaining bicycle quantity in j-th of time period t ime website；
SiRepresent the environment attribute of i-th day；
D represents the festivals or holidays attribute on the same day, and d value is
W represents weather conditions, and w value is
T represents temperature, and t value is
Step (2) K-means is clustered：
Ndate=[n1,n2,...,nj,...,nM] data tuple being concentrated as data sample to be sorted；From data sample
This concentration randomly selects the individual points of k ' as initial classes center, uses μ respectively1,μ2,...,μk′Represent class c1,c2,...,ck′Class in
The heart；Wherein, k ' is positive integer；
K-means clustering algorithms comprise the following steps：
2-1. calculates remaining data tuple Ni′With class center μk′Distance, by remaining data tuple Ni′Assign to and its distance
In most short class；
2-2. recalculates all kinds of new class center μk′ *；
2-3. repeat steps 2-1 and step 2-2, until each data tuple generic all no longer changes；
Step (3) builds Bayes classifier：After classifying using k-means clustering algorithms to set of data samples, k are obtained
Class, each data tuple that data sample is concentrated will obtain a class label；Bayes Method is based on Bayes' theorem：
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mo>|</mo>
<mi>X</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>X</mi>
<mo>|</mo>
<mi>H</mi>
<mo>)</mo>
</mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>X</mi>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
Calculate P (X), P (H) and P (X | H)；
The adaptive PSO Neural Network model training of step (4)
Using sorted data sample training affiliated adaptive PSO Neural Network model of all categories, neutral net is set
Input layer unit number be n, imply layer unit number be p, output layer unit number is q；Then need to adjust in neutral net
Weights number N=n × p+p × q, with the dimension in weights number N problem of representation space for needing to adjust；With vectorial Xi=[xi1,
xi2,...,xiN] a possible value of all weights of neutral net is represented, also illustrate that the position of particle in particle swarm optimization algorithm
Put；PSO Neural Network model training step is as follows：
The position of each particle and speed in 4-1. initialization population populations, the position of particle and speed are between [0,1], i.e.,
Random G vectorial X of generationi=[xi1,xi2,...,xiN], positions of the 1≤i≤G as particle, xiNBetween [0,1]；Again with
G vectorial V of machine generationi=[vi1,vi2,...,viN], speed of the G+1≤i≤2 × G as particle, viNBetween [0,1]；
Set maximum iteration itermax, current iteration number k=1；
4-2. calculates the fitness value of each particle, and it is itself optimal value P to store the current position of each particlebi, store all particles
The minimum position of middle fitness value is global optimum Pbg；
4-3. judges whether k reaches maximum iteration itermax, if so, going to step 4-7, otherwise goes to 4-4；
4-4. updates each particle rapidity and position, calculates the new fitness value of each particle；If the current fitness value of i-th of particle
Better than Pbi, then P is updatedbi；If the optimal location of current particle group is better than Pbg, then P is updatedbg；K=k+1 is set；
4-5. updates inertia weight according to inertia weight calculation formula；
If 4-6. global optimums position PbgDo not change in 10 iteration, then go to step 4-7, otherwise go to step 4-4；
4-7. exports PbgWeights as the PSO Neural Network trained；
The calculation formula and variable implication used in step (4) is as follows：
Fitness value fitness (Xg) calculation formula be：
<mrow>
<mi>f</mi>
<mi>i</mi>
<mi>t</mi>
<mi>n</mi>
<mi>e</mi>
<mi>s</mi>
<mi>s</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>X</mi>
<mi>g</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mn>2</mn>
<msup>
<mi>M</mi>
<mo>&prime;</mo>
</msup>
</mrow>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>l</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<msup>
<mi>M</mi>
<mo>&prime;</mo>
</msup>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>q</mi>
</munderover>
<mrow>
<mo>(</mo>
<mo>(</mo>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>p</mi>
</munderover>
<msub>
<mi>w</mi>
<mrow>
<mi>j</mi>
<mi>k</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>+</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msub>
<mi>w</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msub>
<mi>x</mi>
<mrow>
<mi>l</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>h</mi>
<mi>j</mi>
</mrow>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>o</mi>
<mi>k</mi>
</mrow>
</msub>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>d</mi>
<mrow>
<mi>l</mi>
<mi>k</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
M ' is that data sample concentrates data tuple number；
xliFor i-th of input of l-th of data sample；
θhjFor the threshold value of j-th of unit of hidden layer；
θokFor the threshold value of k-th of unit of output layer；
wijFor the connection weight of i-th of input layer and j-th of hidden layer；
wjkFor the connection weight of j-th of hidden layer and k-th of output layer；
dlkFor k-th of output of l-th of data sample；
Speed calculation formula is：
<mrow>
<msubsup>
<mi>V</mi>
<mi>i</mi>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mo>=</mo>
<mi>w</mi>
<mo>&times;</mo>
<msubsup>
<mi>V</mi>
<mi>i</mi>
<mi>k</mi>
</msubsup>
<mo>+</mo>
<msub>
<mi>c</mi>
<mn>1</mn>
</msub>
<mo>&times;</mo>
<msub>
<mi>r</mi>
<mn>1</mn>
</msub>
<mo>&times;</mo>
<mrow>
<mo>(</mo>
<msup>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
<mi>k</mi>
</msup>
<mo>-</mo>
<msup>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mi>k</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msub>
<mi>c</mi>
<mn>2</mn>
</msub>
<mo>&times;</mo>
<msub>
<mi>r</mi>
<mn>2</mn>
</msub>
<mo>&times;</mo>
<mrow>
<mo>(</mo>
<msup>
<msub>
<mi>P</mi>
<mi>g</mi>
</msub>
<mi>k</mi>
</msup>
<mo>-</mo>
<msup>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mi>k</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
Vi kFor particle XiIn the speed of kth time iteration；
r1,r2It is random number of the value between [0,1]；
c1, c2For one group of random number, c is taken1=c2=2.5；
ω is inertia weight；
Inertia weight ωkCalculation formula is：
<mrow>
<msub>
<mi>&omega;</mi>
<mi>k</mi>
</msub>
<mo>=</mo>
<mi>&omega;</mi>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
<mo>-</mo>
<mfrac>
<mrow>
<mi>&omega;</mi>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
<mo>-</mo>
<mi>&omega;</mi>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mrow>
<mi>i</mi>
<mi>t</mi>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</mfrac>
<mo>&times;</mo>
<mi>k</mi>
</mrow>
2
ω max=0.9, ω min=0.2；
The position calculation formula of particle is：
<mrow>
<msup>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msup>
<mo>=</mo>
<msup>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mi>k</mi>
</msup>
<mo>+</mo>
<msup>
<msub>
<mi>V</mi>
<mi>i</mi>
</msub>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msup>
<mo>;</mo>
</mrow>
Xi kThe position of i-th of particle during iteration secondary for kth；
Step (5) leases model prediction：According to the environment attribute vector S to be predicted of inputpreWith the P (H) being calculated, P (X |
H value), the probability that the data tuple belongs to of all categories, the maximum classification of output posterior probability are calculated；
Step (6) public bicycles lease prediction
The classification exported in model prediction is leased according to step (5), selects corresponding adaptive PSO Neural Network model, will
The remaining public bicycles quantity of preceding n ' individual periods inputs the model, and output is the public bicycles quantity of prediction.
2. a kind of public bicycles based on multisource data fusion as claimed in claim 1 lease Forecasting Methodology, its feature exists
Calculation formula is as follows described in step (2)：
N in 2-1i′With class center μk′Distance calculation formula it is as follows：
<mrow>
<msub>
<mi>D</mi>
<mrow>
<msub>
<mi>N</mi>
<msup>
<mi>i</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>,</mo>
<msub>
<mi>c</mi>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
</msub>
</mrow>
</msub>
<mo>=</mo>
<msqrt>
<mrow>
<mi>&Sigma;</mi>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>N</mi>
<msup>
<mi>i</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
New class center calculation formula is as follows in 2-2：
<mrow>
<msup>
<msub>
<mi>&mu;</mi>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>*</mo>
</msup>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<msup>
<mi>i</mi>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mrow>
<mo>|</mo>
<msub>
<mi>n</mi>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>|</mo>
</mrow>
</munderover>
<msub>
<mi>N</mi>
<mrow>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
<msup>
<mi>i</mi>
<mo>&prime;</mo>
</msup>
</mrow>
</msub>
</mrow>
<mrow>
<mo>|</mo>
<msub>
<mi>n</mi>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
</mrow>
μk′ *For class ck′New class center；
|nk′| it is class ck′The number of middle data tuple；
Nk′i′For class ck′In the i-th ' individual data tuple.
3. a kind of public bicycles based on multisource data fusion as claimed in claim 1 lease Forecasting Methodology, its feature exists
The calculation formula and variable implication used in step (3) is as follows：
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>H</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<msub>
<mi>c</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>T</mi>
</mrow>
</msub>
<mo>|</mo>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
</mrow>
|ck,T| it is to belong to class c in set of data samples TkSample number；
L is the total sample number in set of data samples T；
P (X | H)=P (di|ck)P(wi|ck)P(ti|ck),
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>d</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mi>c</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<msub>
<mi>d</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msub>
<mi>c</mi>
<mi>k</mi>
</msub>
</mrow>
</msub>
<mo>|</mo>
</mrow>
<mrow>
<mo>|</mo>
<msub>
<mi>c</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>T</mi>
</mrow>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
<mo>,</mo>
</mrow>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>w</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mi>c</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<msub>
<mi>w</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msub>
<mi>c</mi>
<mi>k</mi>
</msub>
</mrow>
</msub>
<mo>|</mo>
</mrow>
<mrow>
<mo>|</mo>
<msub>
<mi>c</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>T</mi>
</mrow>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
</mrow>
<mrow>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>t</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mi>c</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<mo>|</mo>
<msub>
<mi>t</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msub>
<mi>c</mi>
<mi>k</mi>
</msub>
</mrow>
</msub>
<mo>|</mo>
</mrow>
<mrow>
<mo>|</mo>
<msub>
<mi>c</mi>
<mrow>
<mi>k</mi>
<mo>,</mo>
<mi>T</mi>
</mrow>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
</mrow>
For class ckMiddle attribute d value, it is diNumber of samples；
For class ckMiddle attribute w value, it is wiNumber of samples；
For class ckMiddle attribute t value, it is tiNumber of samples.
4. a kind of public bicycles based on multisource data fusion as claimed in claim 1 lease Forecasting Methodology, its feature exists
The calculation formula used in step (6) is as follows：
Neural network model output O is specific as follows：
<mrow>
<mi>O</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>p</mi>
</munderover>
<msub>
<mi>w</mi>
<mrow>
<mi>j</mi>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>/</mo>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>+</mo>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<msup>
<mi>n</mi>
<mo>&prime;</mo>
</msup>
</munderover>
<msub>
<mi>w</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msub>
<mi>x</mi>
<mrow>
<mi>l</mi>
<mi>i</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>h</mi>
<mi>j</mi>
</mrow>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mrow>
<mi>o</mi>
<mi>k</mi>
</mrow>
</msub>
<mo>.</mo>
</mrow>
4
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510154943.4A CN104778508B (en) | 2015-04-02 | 2015-04-02 | A kind of public bicycles based on multisource data fusion lease Forecasting Methodology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510154943.4A CN104778508B (en) | 2015-04-02 | 2015-04-02 | A kind of public bicycles based on multisource data fusion lease Forecasting Methodology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104778508A CN104778508A (en) | 2015-07-15 |
| CN104778508B true CN104778508B (en) | 2017-12-08 |
Family
ID=53619961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510154943.4A Expired - Fee Related CN104778508B (en) | 2015-04-02 | 2015-04-02 | A kind of public bicycles based on multisource data fusion lease Forecasting Methodology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104778508B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107045673B (en) * | 2017-03-31 | 2020-09-29 | 杭州电子科技大学 | Public bicycle flow variation prediction method based on stack model fusion |
| CN107145714B (en) * | 2017-04-07 | 2020-05-22 | 浙江大学城市学院 | Multi-factor-based public bicycle usage amount prediction method |
| CN107038503A (en) * | 2017-04-18 | 2017-08-11 | 广东工业大学 | A kind of Demand Forecast method and system of shared equipment |
| CN107274256A (en) * | 2017-05-17 | 2017-10-20 | 南京昱立信息科技有限公司 | Shared bicycle user identification system and stage division |
| CN107301586B (en) * | 2017-06-09 | 2020-10-27 | 中国联合网络通信集团有限公司 | Rentable vehicle prediction method, rentable vehicle prediction device and server |
| CN107704969A (en) * | 2017-10-18 | 2018-02-16 | 南京邮电大学 | A kind of Forecast of Logistics Demand method based on Weighted naive bayes algorithm |
| CN108960476B (en) * | 2018-03-30 | 2022-04-15 | 山东师范大学 | AP-TI clustering-based shared bicycle flow prediction method and device |
| CN108629522B (en) * | 2018-05-11 | 2020-06-16 | 东南大学 | Public bicycle scheduling method based on cluster analysis |
| CN108876056A (en) * | 2018-07-20 | 2018-11-23 | 广东工业大学 | A kind of shared bicycle Demand Forecast method, apparatus, equipment and storage medium |
| CN109543922B (en) * | 2018-12-20 | 2021-04-20 | 西安电子科技大学 | Time-period borrowing and returning amount prediction method for single-vehicle station group shared by piles |
| CN111126641B (en) * | 2019-11-25 | 2023-08-22 | 泰康保险集团股份有限公司 | Resource allocation method and device |
| CN111144648B (en) * | 2019-12-25 | 2023-11-24 | 中国联合网络通信集团有限公司 | People flow prediction device and method |
| CN111523560B (en) * | 2020-03-18 | 2023-07-25 | 第四范式（北京）技术有限公司 | Method, device and system for training number prediction model of arrival trucks |
| CN116738415A (en) * | 2023-08-10 | 2023-09-12 | 北京中超伟业信息安全技术股份有限公司 | Particle swarm optimization weighted naive Bayesian intrusion detection method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104252653A (en) * | 2013-06-26 | 2014-12-31 | 国际商业机器公司 | Method and system for deploying bicycle between public bicycle stations |
| CN104361398A (en) * | 2014-08-04 | 2015-02-18 | 浙江工业大学 | Method for predicting natural demands on public bicycle rental spots |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101045217B1 (en) * | 2010-02-26 | 2011-06-30 | (주)웨이버스 | Managing method for rental bicycle |
-
2015
- 2015-04-02 CN CN201510154943.4A patent/CN104778508B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104252653A (en) * | 2013-06-26 | 2014-12-31 | 国际商业机器公司 | Method and system for deploying bicycle between public bicycle stations |
| CN104361398A (en) * | 2014-08-04 | 2015-02-18 | 浙江工业大学 | Method for predicting natural demands on public bicycle rental spots |
Non-Patent Citations (2)
| Title |
|---|
| Public Bicycle Traffic Flow Prediction based on a Hybrid Model;Haitao Xu et al.;《Applied Mathematics & Information Sciences》;20130301;第7卷(第2期);全文 * |
| 基于混沌粒子群优化小波神经网络的短时交通流预测;沈永增;《计算机应用软件》;20140615;第31卷(第6期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104778508A (en) | 2015-07-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104778508B (en) | A kind of public bicycles based on multisource data fusion lease Forecasting Methodology | |
| Sekhar et al. | Mode choice analysis using random forrest decision trees | |
| Xu et al. | A hybrid machine learning model for demand prediction of edge-computing-based bike-sharing system using Internet of Things | |
| CN110717627B (en) | Full traffic prediction method based on dual graph framework | |
| CN104657744B (en) | A kind of multi-categorizer training method and sorting technique based on non-determined Active Learning | |
| CN106910199B (en) | Car networking crowdsourcing method towards city space information collection | |
| CN106529818B (en) | Water quality assessment Forecasting Methodology based on Fuzzy Wavelet Network | |
| CN110138595A (en) | Time link prediction technique, device, equipment and the medium of dynamic weighting network | |
| CN106781489A (en) | A kind of road network trend prediction method based on recurrent neural network | |
| Lin et al. | A spatial-temporal hybrid model for short-term traffic prediction | |
| CN103049651A (en) | Method and device used for power load aggregation | |
| CN106777987B (en) | River water cluster and quality evaluation method based on hydrodynamic force, water quality and SOM neural network | |
| Azimi et al. | A novel clustering algorithm based on data transformation approaches | |
| CN107392389A (en) | Taxi dispatching processing method based on ARIMA models | |
| WO2022257190A1 (en) | Quantum walk-based multi-feature simulation method for behavior trajectory sequences | |
| Čertický et al. | Fully agent-based simulation model of multimodal mobility in European cities | |
| CN112071062A (en) | Driving time estimation method based on graph convolution network and graph attention network | |
| CN107220724A (en) | Passenger flow forecast method and device | |
| CN114036135A (en) | Method and system for estimating urban mobile source pollution emission by using incomplete information | |
| CN104199870A (en) | Method for building LS-SVM prediction model based on chaotic search | |
| CN110263860A (en) | A kind of freeway traffic flow prediction technique and device | |
| CN109446420A (en) | A kind of cross-domain collaborative filtering method and system | |
| Chen et al. | H-ConvLSTM-based bagging learning approach for ride-hailing demand prediction considering imbalance problems and sparse uncertainty | |
| Al-Dousari et al. | Land use land cover change detection and urban sprawl prediction for Kuwait metropolitan region, using multi-layer perceptron neural networks (MLPNN) | |
| CN112862084B (en) | Traffic flow prediction method based on deep migration fusion learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20171208 Termination date: 20210402 |