CN107832887A - A kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net - Google Patents
A kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net Download PDFInfo
- Publication number
- CN107832887A CN107832887A CN201711099645.5A CN201711099645A CN107832887A CN 107832887 A CN107832887 A CN 107832887A CN 201711099645 A CN201711099645 A CN 201711099645A CN 107832887 A CN107832887 A CN 107832887A
- Authority
- CN
- China
- Prior art keywords
- input
- receptible
- sample
- output
- car
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000007935 neutral effect Effects 0.000 title claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 61
- 238000013528 artificial neural network Methods 0.000 claims abstract description 45
- 210000002569 neuron Anatomy 0.000 claims description 29
- 201000004792 malaria Diseases 0.000 claims description 15
- 238000013210 evaluation model Methods 0.000 claims description 7
- 201000004569 Blindness Diseases 0.000 abstract description 3
- 238000011156 evaluation Methods 0.000 abstract description 2
- 239000013598 vector Substances 0.000 description 27
- 230000008569 process Effects 0.000 description 11
- 238000013507 mapping Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 230000000644 propagated effect Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000007726 management method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 241000001667 Eueretagrotis sigmoides Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003585 interneuronal effect Effects 0.000 description 1
- 210000003127 knee Anatomy 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 238000011045 prefiltration Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 238000011282 treatment Methods 0.000 description 1
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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
- G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman 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
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0621—Item configuration or customization
-
- 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
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0639—Item locations
-
- 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
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0645—Rental transactions; Leasing transactions
-
- 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
- G06Q50/40—Business processes related to the transportation industry
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Accounting & Taxation (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Traffic Control Systems (AREA)
- Navigation (AREA)
Abstract
The present invention relates to a kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net, the training sample pair of setting quantity is obtained first, each training sample is to including input training sample and corresponding output desired value, then BP neural network is established, using each training sample to being iterated training to BP neural network, finally each input parameter to be detected is input in the BP neural network trained, returned the car a position with a pick-up point position, favorite vehicle, optimal drive route, most economical consumption total price and terminal nearest corresponding to obtaining.People's Comprehensive Evaluation trip parameter is helped, decision-making rational routes, so as to reach the purpose sought optimum drive experience, improve vehicle selection mode, effectively prevent the blindness for selecting car and trip, randomness.
Description
Technical field
The present invention relates to a kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net.
Background technology
In recent years, as people are to the pay attention to day by day of environmental protection concept, using electric energy as power resources electric automobile just by
The favor of countries in the world, zero-emission, no pollution, performance brilliance have become its main feature, it was predicted that following electric automobile
Application market still has the development advanced by leaps and bounds.It is supporting with it from present circumstances although electric automobile is with the obvious advantage
Electrically-charging equipment quantity is very few, the matching imperfection between knee, charging security, vehicle is expensive, parking stall management is chaotic,
The problems such as new energy licence plate is rare, it is to restrict electric automobile to popularize the key factor used on a large scale among common citizen.
In order to allow more citizen to enjoy the facility of electric automobile, while do not perplexed again by above mentioned problem, mainstream vendor
The shared electric automobile charter business released, has become the new selection of go off daily.But people are in face of different pick-up
Point, different automobile types, different monovalent, different comfort levels shared electric automobile when, it is often at a loss as to what to do, how going out according to oneself
Row plan, chooses oneself favorite vehicle in nearest distance, while with optimal drive route and most economical consumption
Total price arrives at as early as possible, is the problem of needing to consider emphatically.
The content of the invention
It is an object of the invention to provide a kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net.
To achieve the above object, the solution of the present invention includes following technical scheme.
Method scheme one：This programme provides a kind of shared automobile intelligent optimizing decision-making technique based on neutral net, including
Following steps：
(1) training sample pair of setting quantity is obtained, each training sample is to including input training sample and corresponding output
Value, input training sample include following input parameter：Current location, originating location, final position, receptible pick-up point recently
Distance range, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, highest
Speed per hour, minimum speed per hour, comfort level, vehicle, seating capacity and color, output valve include following output parameter：Nearest pick-up point position
Put, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position；
(2) BP neural network is established, using each training sample to being iterated training to BP neural network, to obtain intelligence
Optimizing decision model；
(3) input parameter to be detected is input in intelligent optimizing decision model, to obtain corresponding nearest pick-up
Point position, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position.
Due to the Nonlinear Mapping relation of input and output with height of system, traditional algorithm can not be realized accurately, base
Can solve this problem very well in the learning algorithm of BP neural network, and energy intelligent decision goes out optimal trip parameter.BP nerves
Network can imitate the mode of human brain process problem, including the processes such as the processing to information, processing, storage and search, energy
Enough reliable treatments problem of nonlinear mapping, have very strong robustness and fault-tolerant ability.Separated out using BP neural network total score
Row demand parameter, in combination with vehicle, map and traffic conditions, people's intelligent selection is helped to share automobile, decision-making drive route,
It is the desirable route for solving the problems, such as trip.Moreover, the powerful robustness of BP neural network, fault-tolerant ability and Nonlinear Mapping energy
The advantages such as power, preferably pick up the car point position, vehicle, drive route, consumption total price, the parameters such as a position of returning the car can be exported.
Under the trend that information interconnection and intercommunication and shared automobile develop rapidly, the present invention helps people's Comprehensive Evaluation by a manner of intelligence
Trip parameter, saving seek car time, decision-making rational routes, save Trip Costs, seek optimum drive experience, improvement so as to reach
The purpose of vehicle selection mode, it effectively prevent the blindness for selecting car and trip, randomness.Also, according to specific trip requirements,
The trip planning of optimization more targeted can be learnt, improves traditional trip mode, avoids blindness and randomness.
Method scheme two：On the basis of method scheme one, the amendment sample pair of setting quantity is obtained, it is each to correct sample pair
Including Introduced Malaria sample and corresponding output desired value, Introduced Malaria sample includes following input parameter：Current location, originate
Position, final position, receptible point distance range of picking up the car recently, receptible highest starting fare, every kilometer of highest or every small
When unit price, festivals or holidays competitively priced ratio, F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color, export the phase
Prestige value includes following output parameter：A nearest pick-up point position, favorite vehicle, optimal drive route, most economical disappear
Expense total price and terminal are returned the car a position；Using each amendment sample to being modified to obtained intelligent optimizing decision model, make mould
Error between the real output value of type and corresponding output desired value is in setting range.
Method scheme three：On the basis of method scheme one or two, using each training sample to entering to BP neural network
Before row iteration training, various input parameters are normalized.
Method scheme four：On the basis of method scheme two, the BP neural network includes input layer, hidden layer and output
Layer, sets any neuron of input layer to the connection weight between any neuron of hidden layer as the first connection weight, and setting is hidden
It is the second connection weight containing any neuron of layer to the connection weight between any neuron of output layer, sets any god of hidden layer
Threshold value through member is first threshold, sets the threshold value of any neuron of output layer as Second Threshold；Correcting intelligent optimizing decision-making
In model process, when the real output value and corresponding output desired value that are obtained according to the Introduced Malaria sample of amendment sample centering
Between error when outside setting range, error amount is propagated back into input layer, and correct first connection weight,
Two connection weights, first threshold and Second Threshold, until error is in setting range.
Method scheme five：On the basis of method scheme one or two, receptible taking recently in training sample will be inputted
Car point distance range, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio,
Ten kinds of F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color parameters are divided into four types, are respectively：
Profit evaluation model data, the input parameter being related to are festivals or holidays competitively priced ratio；
Cost type data, the input parameter being related to are receptible point distance range, the receptible highest of picking up the car recently
Starting fare, every kilometer of highest or unit price hourly；
Disordered data, the input parameter being related to are F-Zero, minimum speed per hour and seating capacity；
Qualitative data, the input parameter being related to are comfort level, vehicle and color；
The input parameter of above-mentioned four type is normalized according to corresponding method for normalizing.
System schema one：This programme provides a kind of shared automobile intelligent optimizing decision system based on neutral net, including
For performing the intelligent optimizing decision-making module of following intelligent optimizing decision strategy：
(1) training sample pair of setting quantity is obtained, each training sample is to including input training sample and corresponding output
Value, input training sample include following input parameter：Current location, originating location, final position, receptible pick-up point recently
Distance range, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, highest
Speed per hour, minimum speed per hour, comfort level, vehicle, seating capacity and color, output valve include following output parameter：Nearest pick-up point position
Put, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position；
(2) BP neural network is established, using each training sample to being iterated training to BP neural network, to obtain intelligence
Optimizing decision model；
(3) input parameter to be detected is input in intelligent optimizing decision model, to obtain corresponding nearest pick-up
Point position, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position.
System schema two：On the basis of system schema one, the amendment sample pair of setting quantity is obtained, it is each to correct sample pair
Including Introduced Malaria sample and corresponding output desired value, Introduced Malaria sample includes following input parameter：Current location, originate
Position, final position, receptible point distance range of picking up the car recently, receptible highest starting fare, every kilometer of highest or every small
When unit price, festivals or holidays competitively priced ratio, F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color, export the phase
Prestige value includes following output parameter：A nearest pick-up point position, favorite vehicle, optimal drive route, most economical disappear
Expense total price and terminal are returned the car a position；Using each amendment sample to being modified to obtained intelligent optimizing decision model, make mould
Error between the real output value of type and corresponding output desired value is in setting range.
System schema three：On the basis of system schema one or two, using each training sample to entering to BP neural network
Before row iteration training, various input parameters are normalized.
System schema four：On the basis of system schema two, the BP neural network includes input layer, hidden layer and output
Layer, sets any neuron of input layer to the connection weight between any neuron of hidden layer as the first connection weight, and setting is hidden
It is the second connection weight containing any neuron of layer to the connection weight between any neuron of output layer, sets any god of hidden layer
Threshold value through member is first threshold, sets the threshold value of any neuron of output layer as Second Threshold；Correcting intelligent optimizing decision-making
In model process, when the real output value and corresponding output desired value that are obtained according to the Introduced Malaria sample of amendment sample centering
Between error when outside setting range, error amount is propagated back into input layer, and correct first connection weight,
Two connection weights, first threshold and Second Threshold, until error is in setting range.
System schema five：On the basis of system schema one or two, receptible taking recently in training sample will be inputted
Car point distance range, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio,
Ten kinds of F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color parameters are divided into four types, are respectively：
Profit evaluation model data, the input parameter being related to are festivals or holidays competitively priced ratio；
Cost type data, the input parameter being related to are receptible point distance range, the receptible highest of picking up the car recently
Starting fare, every kilometer of highest or unit price hourly；
Disordered data, the input parameter being related to are F-Zero, minimum speed per hour and seating capacity；
Qualitative data, the input parameter being related to are comfort level, vehicle and color；
The input parameter of above-mentioned four type is normalized according to corresponding method for normalizing.
Brief description of the drawings
Fig. 1 is the shared automobile intelligent optimizing decision system composition frame chart based on neutral net；
Fig. 2 is the topological structure schematic diagram of BP neural network；
Fig. 3 is the learning process schematic diagram of BP neural network.
Embodiment
The present invention will be further described in detail below in conjunction with the accompanying drawings.
According to the various trip parameters of user preset, local shared electric automobile management subsystem, satellite map subsystem
Carry out comprehensive analysis with the data in real-time traffic subsystem, it is expected to obtain nearest pick-up point position, favorite vehicle, optimal
Drive route, most economical consumption total price and terminal return the car position etc. output.Certain be present between this input and output
Kind mapping relations, but it is non-linear with height, it can not often be accurately determined this mapping relations completely by theoretical calculation；
Simultaneously because the influence of human factor and enchancement factor, input data may and actual value deviation be present, this builds for traditional
Mould mode is probably fatal, it is impossible to subsequently to provide reliable model.BP neural network will be fine with its unique advantage
Solve the above problems.
In order to realize the shared electric automobile intelligence optimizing decision-making technique provided by the invention based on neutral net, tie below
The specific system of unification kind is formed to illustrate this method.It is as shown in Figure 1 the block diagram of system of the present invention, mainly including parameter
Preset system, data pretreatment and Processing with Neural Network system.Wherein, preset parameter system calls local shared electronic vapour
Car manages subsystem, and user can carry out parameter according to actual trip requirements and user preferences and preset；Data prediction system
Completely a pair of arrange parameters are classified and normalized；Processing with Neural Network system needs training sample to carry out network self-study
Practise, to optimize network structure and parameter, further, in order to ensure accuracy, it is necessary to carry out net with enough training samples
Network self study, moreover, in order to obtain authentic and valid data, Processing with Neural Network system need to call satellite map subsystem and
Real-time traffic subsystem.Wherein satellite map subsystem is that the data that BP neural network provides have：The current location of user, originate
Position, final position, all optional drive routes, distance total distance, E.T.A etc.；Real-time traffic subsystem is BP
The data that neutral net provides have：The jam situation of every optional drive route, on the way construction or barrier situation, traffic control
Situation, traffic lights number etc..In addition, it can therefrom obtain following number when calling local shared electric automobile management subsystem
According to：1. idle vehicle parameter：Position, number etc.；2. performance parameter：The remaining SOC ratios (or remaining milimeter number) of idle vehicle,
F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity, color etc.；3. billing model parameter：Starting fare, every kilometer it is (or every
Hour) unit price, festivals or holidays competitively priced ratio etc..
The fixed index that user goes on a journey in preset parameter system has：Current location, originating location, final position；Judge refers to
Indicate：Receptible point distance (i.e. walking pick-up distance), receptible highest starting fare, the receptible highest of picking up the car recently is every
Kilometer (or per hour) unit price, festivals or holidays competitively priced ratio, F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity, face
Color etc.., can will if user preferences are relatively more fixed in addition, it is necessary to be configured to each of the above parameter when going on a journey for the first time
Partial parameters use as default, and next time before travel only need to be to remaining parameter setting.
Therefore, from above-mentioned local shared electric automobile management subsystem, satellite map subsystem and real-time traffic subsystem
Following 13 kinds of inputs parameter can be obtained in all data of middle acquisition, is respectively：Current location, originating location, final position,
Receptible point distance (i.e. walking pick-up distance), receptible highest starting fare, receptible every kilometer of the highest of picking up the car recently
(or per hour) unit price, festivals or holidays competitively priced ratio, F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color.
This 13 kinds input parameters are used for the training of follow-up BP neural network.
In addition to fixed index, attributive classification is carried out to the data judged of needs, because preset data is made up of multiple indexs,
There is quantitative target, also difinite quality index, for quantitative target, its property and dimension also have difference.It is bigger for profit evaluation model index
Better；It is for cost type index, then the smaller the better；For interval type index, property value is best in a certain fixed interval.Exactly
Due to each parameter dimension and the difference of property, cause between each index can not the property spent together, thus refer to firstly the need of these
Mark is classified according to attribute.
Therefore, the index for needing to carry out attributive classification in data pretreatment shares 10, can be divided into 4 kinds：1. benefit
Type (more big more excellent type) index：Festivals or holidays competitively priced ratio；Cost type 2. (smaller more excellent type) index：It is receptible to take recently
Car point distance (i.e. walking pick-up distance), receptible highest starting fare, receptible every kilometer of highest (or per hour) are monovalent；
Interval type 3. (being excellent type in a certain fixed interval) index：F-Zero, minimum speed per hour, seating capacity；4. qualitative type index：Comfortably
Degree, vehicle, color.
For above-mentioned all types of several indexs, the vector corresponding to data of all categories given below：
1. the vector of profit evaluation model data composition is set as a, then a=[a1], wherein, a1For festivals or holidays competitively priced ratio.
2. the vector for being set as this type data composition is b, then b=[b1, b2, b3], wherein, b1Picked up the car recently to be receptible
Point distance (i.e. walking pick-up distance), b2For receptible highest starting fare, b3For receptible every kilometer of highest (or per hour)
Unit price.
3. the vector of disordered data composition is set as c, then c=[c1, c2, c3], wherein, c1For F-Zero, c2To be minimum
Speed per hour, c3For seating capacity.
4. the vector of setting property data composition is d, then d=[d1, d2, d3], wherein, d1For vehicle, d2For comfort level, d3For
Color.
To sorted index in data pretreatment, dimension is carried out by corresponding utility function and is mapped to
In one limited section, i.e. normalized.
If before data normalization, input vector corresponding to neural network input layer is R, then R=[a, b, c, d]=[a1, b1,
b2, b3, c1, c2, c3, d1, d2, d3]=[R1, R2, R3, R4, R5, R6, R7, R8, R9, R10], it is seen that each input sample is one
Multi-C vector.Although there is no fixed index in each compositing factor in above-mentioned input vector：Current location, originating location and terminal
Position, but these three fixed indexs are the training for participating in follow-up BP neural network, are equally the inputs of BP neural network
Parameter.
In the present embodiment, in data pretreatment, in order to which will there are different dimensions and the data of attribute after above-mentioned classification,
The dimensionless index value being converted on closed interval [0,1], their maximum r is determined on respective domainmaxAnd minimum value
rminWith average value ravg, then it is normalized by utility function, method is as follows：
1. profit evaluation model data：For arbitrary ai∈ a,
2. cost type data：For arbitrary bi∈ b,
3. disordered data：For arbitrary ci∈ c, work as ci∈[cmin, cavg] when, Work as ci∈[cavg, cmax] when,Work as ciFor
During other values,For 0.
4. qualitative data：For arbitrary di3 attribute are included in ∈ d, vectorial d：Vehicle, comfort level and color, if Q1、
Q2、Q3Corresponding attribute is represented respectively.According to attribute Q1Element in d is ranked up, if by attribute Q1Regard a fuzzy son as
Collection, as long as each element can be obtained in d to Q1Degree of membership, so that it may by qualitative data quantification.Comprise the following steps that：
1) any two element is established in d on Q1It is relatively several to (fy(x), fx(y) 0 ＜ (f), are mety(x), fx
(y)) ＜ 1.Wherein, fy(x) for representing relative y, x has certain attribute Q1Intensity；fx(y) for representing relative x, y has
Certain attribute Q1Intensity.
2) establish with uxy=fy(x)/max(fy(x), fx(y) it is) matrix U=(u of elementxy), it is clear that there is uxy∈ [0,
, and u 1]xx=uyy=uzz=...=1.
3) matrix U=(u is takenxy) in each row degree of membership of the maximum as each row corresponding element, establish in limited domain
On fuzzy subset Q1Membership function.For arbitrary di, dj∈ d, set x=d respectivelyi, y=dj, then diWith respect to Q1Person in servitude
Category degree isSimilarly, diWith respect to Q2、Q3Degree of membership beHave
In data pretreatment, the input vector after data normalization is R*, then have
Remain as multi-C vector.
As shown in Fig. 2 be the topological structure of BP neural network, including input layer, hidden layer and output layer 3-tier architecture, profit
Network training is carried out with feedforward network and error back propagation learning algorithm, there is very strong learning ability, adaptivity, robust
Property and non-linear mapping capability and fault-tolerant ability.In the present embodiment, carried out to input initial data to user at unified
Reason, adds data prediction layer, the prefilter layer as input layer.The general principle of BP neural network is：
(1) net definitions
According to network model, each layer relation of definable：If input layer has m node (neuron), corresponding input vector is X
=(x1, x2..., xm)T；Hidden layer has k node, and corresponding hidden layer output vector is Y=(y1, y2..., yk)T；Output layer
There is n node, corresponding output vector is O=(o1, o2..., on)T.Wherein, m, k, n represent the dimension of vector respectively.
Make input layer any node xiTo hidden layer any node yjBetween connection weight be ωij；Hidden layer node yjArrive
Output layer OlBetween connection weight be ωjl；θjFor the threshold value of j-th of neuron of hidden layer, θlFor l-th of god of network output layer
Threshold value through member, wherein：I=1,2 ..., m, j=1,2 ..., k, l=1,2 ..., n.
Accompanying drawing 2 is compareed, above net definitions are corresponding in the present embodiment, by normalizing pretreated input vector X
=(x1, x2..., xm)TEquivalent toM=10, and three
Fixed index, the specific parameter that inputs are：The current location of user, originating location, final position, receptible pick-up point recently
Distance range, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, highest
Speed per hour, minimum speed per hour, comfort level, vehicle, seating capacity and color；The element of desired output vector includes nearest pick-up point position
Put, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position, map network definition,
There are output vector O=(o respectively1, o2..., on)T, wherein, o1Represent nearest pick-up point position, o2Represent favorite vehicle,
o3Represent optimal drive route, o4Represent most economical consumption total price, o5Represent that terminal is returned the car a position, there is n=5.
Therefore, when being trained to BP neural network, the training sample used is to including two parts：It is input instruction respectively
Practice sample and output valve, each training sample that inputs is corresponding with an output valve, wherein, input training sample is exactly above-mentioned warp
The 10 kinds of input vectors and three fixed indexs crossed after normalized, output valve are above-mentioned 5 kinds of output vectors.
(2) mathematical relationship
If f () is S type excitation functions, i.e. Sigmoid functions：F (x) can continuously be led, and have f,
(x)=f (x) [1-f (x)].
1. for input layer, input exports；
2. for hidden layer, if input vector NetjRepresent, then have：
Input：
Output：
3. for output layer, if input vector NetlRepresent, then have：
Input：
Output：
(3) course of work
When giving BP neural network one input vector, that is, when inputting an input training sample, input signal is through hidden layer
After successively handling, output layer is passed to, and an output vector is produced after being handled by output layer, be i.e. reality output responds, each layer
The state of neuron only under the influence of one layer of neuron state, this is a successively state renewal process, referred to as propagated forward or
Person's forward-propagating.That is, using each training sample to being iterated training to BP neural network, to obtain a model,
Referred to as intelligent optimizing decision model.
Further, in order to ensure the accuracy of model, it is necessary to be modified to model, i.e., when reality output response and phase
When the output valve of prestige has error, then error back propagation is transferred to, i.e., by error amount along the successively backpropagation of original connecting path
Until input layer, and correct each layer connection weight.It is therefore desirable to the amendment sample pair of setting quantity is specially set, each amendment
Sample to including Introduced Malaria sample and corresponding output desired value, Introduced Malaria sample is identical with above-mentioned input training sample,
Including following input parameter：The current location of user, originating location, final position, a receptible point distance range of picking up the car recently,
It is receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, F-Zero, minimum
Speed per hour, comfort level, vehicle, seating capacity and color；Output desired value includes following output parameter：Nearest pick-up point position, most
Vehicle, optimal drive route, most economical consumption total price and the terminal liked are returned the car a position.
For a given amendment sample pair, propagated forward and the process of error back propagation are repeated continuously, is passed through
The interneuronal connection weight of each layer and neuron threshold value are changed on the way so that error reaches minimum, and alternatively referred to as error is being set
In the range of.Also, BP neural network needs to carry out network self study with enough amendment samples, to obtain optimal network
Weights and threshold value.Learning rules are to use steepest descent method, wherein, forward-propagating is used for network calculations, and a certain input is obtained
Its output；Backpropagation is used for successively transmission error, by the iteration of algorithm, constantly adjusts the weights and threshold value of network, makes
The error sum of squares of network constantly reduces, until error meets sets requirement (i.e. in setting range).Therefore, for given
One amendment sample pair, constantly with sample training network one by one, repeated forward is propagated and back-propagation process, when each amendment
When sample is to all meeting to require, illustrate that BP neural network model now has been stablized, obtained the rule hidden in sample, can
With for handling formal sample data.
When each amendment sample is to all meeting to require, illustrate that BP neural network has succeeded in school.
Therefore, each training sample is to by input vector (inputting training sample) and its corresponding output vector group
Into each sample of correcting input vector (i.e. Introduced Malaria sample) and corresponding output desired value to being made up of.Training sample
Pair and amendment sample pair selection it is extremely important to the performance of BP neural network, poor sample is to can not only cause the mistake of network
Mapping relations, and the learning process of network may be made not restrain.Therefore, select the rule of sample pair should for sample
Representative, popularity and compactedness.The sample that the present embodiment is chosen is good to being evaluated from current user, and with representative
Property, popularity and compactedness, each mainstream car design share the complete trip process data of electric automobile.
Below exemplified by correcting sample pair (be equally applicable to training sample to), setting amendment sample is to there is p, then,
Introduced Malaria sample just should mutually have p, be respectively：
S1=(s11, s12..., s1m)T, S2=(s21, s22..., s2m)T..., Sp=(sp1, sp2..., spm)T。
Corresponding output desired value also has p, is respectively：
T1=(t11, t12..., t1n)T, T2=(t21, t22..., t2n)T..., Tp=(tp1, tp2..., tpn)T。
After p-th of Introduced Malaria sample is input to network, by each layer computing, the reality output of network can be finally given
Value Op=(op1, op2..., opn)T, in actual applications, will both it if reality output and corresponding desired output are unequal
Between error be set to E, it is as follows to define the error function：
In formula：tlFor the desired output of network, OlFor reality output.
Error formula is expanded to hidden layer, had：
(for the sake of convenience, often-θlRegard y as0It is constantly equal to corresponding weights when 1)
Further spread out to input layer, have：
From formula (1-6), (1-7), each layer weights ω of network inputs errorjl、ωijFunction, therefore adjust weights
Error E can be changed.
The principle for adjusting weights is error is constantly reduced, so should make the adjustment amount of weights and the negative gradient of error into just
Than that is,：
Negative sign represents that gradient declines in above formula, and constant η ∈ (0,1) be Studying factors, general span be [0.01~
1], the pace of learning of neutral net, therefore also referred to as learning rate are reacted.
In order to from which further follow that clear and definite weighed value adjusting formula, according to the chain type of differential rule, the hidden of network can be derived
Correction containing connection weight between layer and output layer is：
Δμij=η δjOi (1-10)
If order：Then have：
Formula (1-11) is referred to as output layer equivalent error signal, then Δ ωjl=-η δl·Oj, output neuron actual error
For：Δ l=tl-Ol.It can similarly obtain：
If orderThen it is called the error signal of hidden layer.
As can be seen here, the error signal of output layer is relevant with the desired output of network and the difference of reality output, and each implicit
The error signal of layer and the error signal of preceding layers are all relevant, are that successively anti-pass comes since output layer.In summary, respectively
Layer weights adjustment formula be：
As described in accompanying drawing 3, a kind of specific algorithm flow chart of BP neural network is provided, implementation process is as follows：
1. initialize
Setting to connection weight (including threshold value) initial value, the initial value of each connection weight are set to random number；Learning rate η and
Network training error precision E is set to the decimal between (0,1).
2. forward direction calculates each layer output
Training sample pair is inputted, calculates each sample forward sequence the output of each hidden layer, output layer neuron.
3. each layer error signal of backwards calculation
To all amendment samples to (Sp, Tp), calculation error E, and respectively since output layer until input layer, by
The equivalent error δ of layer each layer neuron of backwards calculationi、δj。
4. adjust the connection weight of each layer
According to modified weight formula and above-mentioned δi、δj, each layer connection weight is changed, to reduce error E.
5. return to step is 2., positive calculating is carried out according to new connection weight.If to each amendment sample to (Sp, Tp), its
Output error E is satisfied by required precision (i.e. error is in setting range), then training terminates.Otherwise compute repeatedly, until meeting
Untill it is required that.
6. study terminates.
In the present embodiment, after the completion of BP neural network study, according to algorithm flow, after ultimately forming amendment
Model.In actual applications, actual input parameter is X=(x1, x2..., xm)T, equivalent toM=10, and three fixed indexs.I.e. in practical application
In, the shared electric automobile of reality is hired a car user's input parameter, i.e.,：The current location of user, originating location, final position,
Festivals or holidays competitively priced ratio, it is receptible recently pick up the car point distance (i.e. walking pick-up distance), receptible highest starting fare,
Receptible every kilometer of highest (or per hour) unit price, F-Zero, minimum speed per hour, seating capacity, vehicle, comfort level and color are defeated
Enter into the BP neural network trained, corresponding output valve can be obtained, i.e. output layer output parameter vector is O=(o1,
o2..., on)T, n=5, that is, obtain：A nearest pick-up point position, favorite vehicle, optimal drive route, most economical disappear
Expense total price and terminal are returned the car a position.That is, after BP neural network is handled 13 indexs of input, final is defeated
Go out parameter for nearest pick-up point position, favorite vehicle, optimal drive route, most economical consumption total price and terminal
Return the car a position.
Specific embodiment is presented above, but the present invention is not limited to described embodiment.The base of the present invention
This thinking is above-mentioned basic scheme, for those of ordinary skill in the art, according to the teachings of the present invention, designs various changes
The model of shape, formula, parameter simultaneously need not spend creative work, such as：Technical scheme is not suitable only for electric automobile,
If other kinds of vehicles such as shared fuel-engined vehicles, also within the scope of the present invention.The present invention is not being departed from
Principle and spirit in the case of to embodiment carry out change, modification, replacement and modification still fall within the present invention protection model
In enclosing.
Claims (10)
1. a kind of shared automobile intelligent optimizing decision-making technique based on neutral net, it is characterised in that comprise the following steps：
(1) training sample pair of setting quantity is obtained, each training sample is defeated to including input training sample and corresponding output valve
Entering training sample includes following input parameter：Current location, originating location, final position, receptible point distance of picking up the car recently
Scope, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, F-Zero,
Minimum speed per hour, comfort level, vehicle, seating capacity and color, output valve include following output parameter：Nearest pick-up point position, most
Vehicle, optimal drive route, most economical consumption total price and the terminal liked are returned the car a position；
(2) BP neural network is established, using each training sample to being iterated training to BP neural network, to obtain intelligent optimizing
Decision model；
(3) input parameter to be detected is input in intelligent optimizing decision model, to obtain corresponding nearest pick-up point position
Put, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position.
2. the shared automobile intelligent optimizing decision-making technique according to claim 1 based on neutral net, it is characterised in that obtain
The amendment sample pair of setting quantity is taken, each sample of correcting including Introduced Malaria sample and corresponding output desired value, input to repairing
Positive sample includes following input parameter：Current location, originating location, final position, it is receptible recently pick up the car point a distance range,
It is receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, F-Zero, minimum
Speed per hour, comfort level, vehicle, seating capacity and color, output desired value include following output parameter：Nearest pick-up point position, most
Vehicle, optimal drive route, most economical consumption total price and the terminal liked are returned the car a position；Using each amendment sample to right
Obtained intelligent optimizing decision model is modified, and makes the error between the real output value of model and corresponding output desired value
In setting range.
3. the shared automobile intelligent optimizing decision-making technique according to claim 1 or 2 based on neutral net, its feature exist
In to before being iterated training to BP neural network, place is being normalized to various input parameters using each training sample
Reason.
4. the shared automobile intelligent optimizing decision-making technique according to claim 2 based on neutral net, it is characterised in that institute
Stating BP neural network includes input layer, hidden layer and output layer, the setting any neuron of input layer to any neuron of hidden layer
Between connection weight be the first connection weight, the setting any neuron of hidden layer is to the connection between any neuron of output layer
Weights are the second connection weight, set the threshold value of any neuron of hidden layer as first threshold, set any neuron of output layer
Threshold value be Second Threshold；During intelligent optimizing decision model is corrected, when the Introduced Malaria sample according to amendment sample centering
Originally the error between the real output value obtained and corresponding output desired value reversely passes error amount when outside setting range
Input layer is cast to, and corrects first connection weight, the second connection weight, first threshold and Second Threshold, until error exists
In setting range.
5. the shared automobile intelligent optimizing decision-making technique according to claim 1 or 2 based on neutral net, its feature exist
In receptible point distance range, receptible highest starting fare, every kilometer of the highest of picking up the car recently that will be inputted in training sample
Or unit price hourly, festivals or holidays competitively priced ratio, F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color
Ten kinds of parameters are divided into four types, are respectively：
Profit evaluation model data, the input parameter being related to are festivals or holidays competitively priced ratio；
Cost type data, the input parameter being related to are receptible pick up the car recently point distance range, the starting of receptible highest
Valency, every kilometer of highest or unit price hourly；
Disordered data, the input parameter being related to are F-Zero, minimum speed per hour and seating capacity；
Qualitative data, the input parameter being related to are comfort level, vehicle and color；
The input parameter of above-mentioned four type is normalized according to corresponding method for normalizing.
6. a kind of shared automobile intelligent optimizing decision system based on neutral net, it is characterised in that including following for performing
The intelligent optimizing decision-making module of intelligent optimizing decision strategy：
(1) training sample pair of setting quantity is obtained, each training sample is defeated to including input training sample and corresponding output valve
Entering training sample includes following input parameter：Current location, originating location, final position, receptible point distance of picking up the car recently
Scope, receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, F-Zero,
Minimum speed per hour, comfort level, vehicle, seating capacity and color, output valve include following output parameter：Nearest pick-up point position, most
Vehicle, optimal drive route, most economical consumption total price and the terminal liked are returned the car a position；
(2) BP neural network is established, using each training sample to being iterated training to BP neural network, to obtain intelligent optimizing
Decision model；
(3) input parameter to be detected is input in intelligent optimizing decision model, to obtain corresponding nearest pick-up point position
Put, favorite vehicle, optimal drive route, most economical consumption total price and terminal are returned the car a position.
7. the shared automobile intelligent optimizing decision system according to claim 6 based on neutral net, it is characterised in that obtain
The amendment sample pair of setting quantity is taken, each sample of correcting including Introduced Malaria sample and corresponding output desired value, input to repairing
Positive sample includes following input parameter：Current location, originating location, final position, it is receptible recently pick up the car point a distance range,
It is receptible highest starting fare, every kilometer of highest or unit price hourly, festivals or holidays competitively priced ratio, F-Zero, minimum
Speed per hour, comfort level, vehicle, seating capacity and color, output desired value include following output parameter：Nearest pick-up point position, most
Vehicle, optimal drive route, most economical consumption total price and the terminal liked are returned the car a position；Using each amendment sample to right
Obtained intelligent optimizing decision model is modified, and makes the error between the real output value of model and corresponding output desired value
In setting range.
8. the shared automobile intelligent optimizing decision system based on neutral net according to claim 6 or 7, its feature exist
In to before being iterated training to BP neural network, place is being normalized to various input parameters using each training sample
Reason.
9. the shared automobile intelligent optimizing decision system according to claim 7 based on neutral net, it is characterised in that institute
Stating BP neural network includes input layer, hidden layer and output layer, the setting any neuron of input layer to any neuron of hidden layer
Between connection weight be the first connection weight, the setting any neuron of hidden layer is to the connection between any neuron of output layer
Weights are the second connection weight, set the threshold value of any neuron of hidden layer as first threshold, set any neuron of output layer
Threshold value be Second Threshold；During intelligent optimizing decision model is corrected, when the Introduced Malaria sample according to amendment sample centering
Originally the error between the real output value obtained and corresponding output desired value reversely passes error amount when outside setting range
Input layer is cast to, and corrects first connection weight, the second connection weight, first threshold and Second Threshold, until error exists
In setting range.
10. the shared automobile intelligent optimizing decision system based on neutral net according to claim 6 or 7, its feature exist
In receptible point distance range, receptible highest starting fare, every kilometer of the highest of picking up the car recently that will be inputted in training sample
Or unit price hourly, festivals or holidays competitively priced ratio, F-Zero, minimum speed per hour, comfort level, vehicle, seating capacity and color
Ten kinds of parameters are divided into four types, are respectively：
Profit evaluation model data, the input parameter being related to are festivals or holidays competitively priced ratio；
Cost type data, the input parameter being related to are receptible pick up the car recently point distance range, the starting of receptible highest
Valency, every kilometer of highest or unit price hourly；
Disordered data, the input parameter being related to are F-Zero, minimum speed per hour and seating capacity；
Qualitative data, the input parameter being related to are comfort level, vehicle and color；
The input parameter of above-mentioned four type is normalized according to corresponding method for normalizing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711099645.5A CN107832887A (en) | 2017-11-09 | 2017-11-09 | A kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711099645.5A CN107832887A (en) | 2017-11-09 | 2017-11-09 | A kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107832887A true CN107832887A (en) | 2018-03-23 |
Family
ID=61654069
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711099645.5A Pending CN107832887A (en) | 2017-11-09 | 2017-11-09 | A kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107832887A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110599301A (en) * | 2019-09-05 | 2019-12-20 | 上海钧正网络科技有限公司 | Vehicle management method, device, computer equipment and storage medium |
| CN110992538A (en) * | 2019-12-16 | 2020-04-10 | 上海钧正网络科技有限公司 | Vehicle voice unlocking method, device and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102436598A (en) * | 2011-09-08 | 2012-05-02 | 崔志明 | Optimum vehicle path mining method based on traffic image information |
| CN105139301A (en) * | 2015-08-13 | 2015-12-09 | 厦门维途信息技术有限公司 | Guiding method based on BP neural network |
| CN107238393A (en) * | 2017-06-14 | 2017-10-10 | 赵宇航 | It is a kind of to be gone on a journey Intelligent planning method based on shared economic personnel |
| CN107289958A (en) * | 2017-06-01 | 2017-10-24 | 北京嘀嘀无限科技发展有限公司 | Route planning method, device, server, client, computer-readable storage medium |
-
2017
- 2017-11-09 CN CN201711099645.5A patent/CN107832887A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102436598A (en) * | 2011-09-08 | 2012-05-02 | 崔志明 | Optimum vehicle path mining method based on traffic image information |
| CN105139301A (en) * | 2015-08-13 | 2015-12-09 | 厦门维途信息技术有限公司 | Guiding method based on BP neural network |
| CN107289958A (en) * | 2017-06-01 | 2017-10-24 | 北京嘀嘀无限科技发展有限公司 | Route planning method, device, server, client, computer-readable storage medium |
| CN107238393A (en) * | 2017-06-14 | 2017-10-10 | 赵宇航 | It is a kind of to be gone on a journey Intelligent planning method based on shared economic personnel |
Non-Patent Citations (1)
| Title |
|---|
| 郭桂林: "基于用户选择的多约束条件动态最优路径规划", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110599301A (en) * | 2019-09-05 | 2019-12-20 | 上海钧正网络科技有限公司 | Vehicle management method, device, computer equipment and storage medium |
| CN110599301B (en) * | 2019-09-05 | 2022-03-18 | 上海钧正网络科技有限公司 | Vehicle management method, device, computer equipment and storage medium |
| CN110992538A (en) * | 2019-12-16 | 2020-04-10 | 上海钧正网络科技有限公司 | Vehicle voice unlocking method, device and medium |
| CN110992538B (en) * | 2019-12-16 | 2022-05-27 | 上海钧正网络科技有限公司 | Vehicle voice unlocking method, device and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Prediction for tourism flow based on LSTM neural network | |
| Shang et al. | A unified framework for multicriteria evaluation of transportation projects | |
| Mungle et al. | A fuzzy clustering-based genetic algorithm approach for time–cost–quality trade-off problems: A case study of highway construction project | |
| CN108027899A (en) | Method for the performance for improving housebroken machine learning model | |
| CN106529818A (en) | Water quality evaluation prediction method based on fuzzy wavelet neural network | |
| Ma et al. | A deep reinforcement learning approach to traffic signal control with temporal traffic pattern mining | |
| CN102819769A (en) | Method for optimizing personalized learning resources based on integrated modeling | |
| CN108510126A (en) | A kind of Predictive Methods of Road Accidents based on PCA and BP neural network | |
| CN110399487A (en) | A kind of file classification method, device, electronic equipment and storage medium | |
| CN106355346A (en) | Method and system for predicting box-office performance of movies on basis of neural network algorithms | |
| CN113407864A (en) | Group recommendation method based on mixed attention network | |
| CN107944698A (en) | The manufacture demand of facing cloud manufacture and capacity of equipment normalization modeling method | |
| Luo et al. | Dynamic taxi service planning by minimizing cruising distance without passengers | |
| CN107832887A (en) | A kind of shared automobile intelligent optimizing decision-making technique and system based on neutral net | |
| Huo et al. | Cooperative control for multi-intersection traffic signal based on deep reinforcement learning and imitation learning | |
| CN106779346A (en) | A kind of Forecasting Methodology of monthly power consumption | |
| CN105844334A (en) | Radial basis function neural network-based temperature interpolation algorithm | |
| CN106097094A (en) | A kind of man-computer cooperation credit evaluation new model towards medium-sized and small enterprises | |
| Zhao et al. | A driving risk prediction algorithm based on PCA-BP neural network in vehicular communication | |
| Zhao et al. | Learning transformer-based cooperation for networked traffic signal control | |
| Ming et al. | Cooperative modular reinforcement learning for large discrete action space problem | |
| CN115600421A (en) | Construction method and device and medium of autonomous traffic system evolution model based on improved Petri network | |
| AU2021106177A4 (en) | Method for predicting spatial-temporal dynamic distribution of electric vehicle charging loads | |
| Liu et al. | Co-prediction of multimodal transportation demands with self-learned spatial dependence | |
| CN109859087A (en) | A kind of intelligent Community soft environment configuration system and method surpassing brain using city |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180323 |
|
| RJ01 | Rejection of invention patent application after publication |