CN110097320A - Order forecast method and device, logistics system and computer-readable medium - Google Patents
Order forecast method and device, logistics system and computer-readable medium Download PDFInfo
- Publication number
- CN110097320A CN110097320A CN201910382417.1A CN201910382417A CN110097320A CN 110097320 A CN110097320 A CN 110097320A CN 201910382417 A CN201910382417 A CN 201910382417A CN 110097320 A CN110097320 A CN 110097320A
- Authority
- CN
- China
- Prior art keywords
- order
- data
- history
- route
- order 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000012549 training Methods 0.000 claims abstract description 65
- 230000007787 long-term memory Effects 0.000 claims abstract description 31
- 238000012360 testing method Methods 0.000 claims abstract description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 17
- 230000015654 memory Effects 0.000 claims description 7
- 238000010586 diagram Methods 0.000 description 15
- 238000004891 communication Methods 0.000 description 12
- 230000001537 neural effect Effects 0.000 description 7
- 238000005457 optimization Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000000052 comparative effect Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000006403 short-term memory Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 230000003203 everyday effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002354 daily effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—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
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Human Resources & Organizations (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of order forecast methods and device, logistics system and computer-readable medium.This method comprises: the History Order data of acquisition route, the History Order data include that order generates time and order volume；The History Order data are integrated according to the order generation time；From the N days order volumes of History Order data cutout through integrating as timing list entries, and using the N+1 days order volumes as the label of the timing list entries, mobile data intercepts window and obtains multiple training samples, and the training sample includes timing list entries and label；Use the multiple training sample training shot and long term memory models；The order volume of testing time is predicted using housebroken shot and long term memory models.
Description
Technical field
The invention mainly relates to Intelligent logistics field more particularly to a kind of order forecast method and device, logistics system and
Computer-readable medium.
Background technique
Intelligent logistics transport field is the crossing domain of artificial intelligence technology and logistics field, it is intended to utilize artificial intelligence skill
Intelligent algorithm in art substitutes artificial method to solve the FAQs in logistics field, such as order forecasting problem, path
Planning problem, road junction Plan Problem, vehicle dispatching problem, Warehouse Location problem etc..Wherein order forecasting problem again can basis
Predetermined period is divided into a day degree, weekly, monthly, annual order forecasting etc..It, can be further according to delivery in order forecasting problem
Ground and place of acceptance are divided into each route order forecasting problem, are divided into each vehicle order forecasting problem and more complicated according to vehicle
Order forecasting problem etc. of each vehicle route on the multicycle.
Intelligent logistics transport field solves the problems, such as the scheme of order forecasting, usually empirically formula, is gone through by analysis
History order data, vehicle factor yield, transport power data, market situation etc. carry out linear fit to History Order data, and are solved
Order forecasting problem certainly is mostly monthly or more macrocyclic prediction.This scheme has the following disadvantages:
1, the prediction mode precision of prediction of linear fit is low, response speed is slow, is unable to satisfy the prediction of daily Logistic Scheduling
Demand.
2, higher to the skill requirement of prognosticator, prediction effect is related to the experience height of people, it is difficult to meet intelligent object
Flow the needs of transport field Continuous optimization algorithm.
Summary of the invention
The technical problem to be solved in the present invention is to provide order forecast methods and device, object based on shot and long term memory models
Streaming system and computer storage medium.
In order to solve the above technical problems, an aspect of of the present present invention provides a kind of order forecast method, this method comprises: adopting
Collect the History Order data of route, the History Order data include that order generates time and order volume；It is raw according to the order
The History Order data are integrated at the time；From the N days order volumes of History Order data cutout through integrating as when
Sequence list entries, and using the N+1 days order volumes as the label of the timing list entries, it is more that mobile data intercepts window acquisition
A training sample, the training sample include timing list entries and label；Use the multiple training sample training shot and long term
Memory models；The order volume of testing time is predicted using housebroken shot and long term memory models.
In one embodiment of this invention, the History Order data for acquiring a plurality of route calculate in a plurality of route every two
The similarity of route History Order data merges the two lines road that similarity is greater than preset value.
In one embodiment of this invention, using manhatton distance similarity algorithm, Euclidean distance similarity algorithm, cosine
Similarity algorithm or Pearson's similarity algorithm calculate the similarity of every two lines road History Order data in a plurality of route.
In one embodiment of this invention, according to the order generate the time number of weeks to the History Order data into
Row integration.
In one embodiment of this invention, the History Order data that the step-length of mobile data interception window is Dan Tian.
In one embodiment of this invention, the History Order data for acquiring route include: to obtain the corresponding original of History Order
Beginning data；Initial data is pre-processed, the order for obtaining History Order data generates time and order volume.
In one embodiment of this invention, the pretreatment includes at least one in outlier processing or missing values processing
Kind.
It in one embodiment of this invention, further include the online data for obtaining order, simultaneously based on online data training
Update the shot and long term memory models.
Another aspect provides a kind of order forecasting device, which includes: acquisition unit, acquires route
History Order data, the History Order data include that order generates time and order volume；Integral unit, it is raw according to the order
The History Order data are integrated at the time；Training sample acquiring unit, from the History Order data cutout N through integrating
It order volume moves number as timing list entries, and using the N+1 days order volumes as the label of the timing list entries
Multiple training samples are obtained according to interception window, the training sample includes timing list entries and label；Training unit uses institute
State multiple training sample training shot and long term memory models；Predicting unit, when using housebroken shot and long term memory models to test
Between order volume predicted.
In one embodiment of this invention, the acquisition unit acquires the History Order data of a plurality of route, calculates a plurality of
The similarity of every two lines road History Order data in route merges the two lines road that similarity is greater than preset value.
In one embodiment of this invention, the acquisition unit uses manhatton distance similarity algorithm, Euclidean distance phase
Every two lines road History Order data in a plurality of route are calculated like degree algorithm, cosine similarity algorithm or Pearson's similarity algorithm
Similarity.
In one embodiment of this invention, the integral unit is gone through according to the number of weeks that the order generates the time to described
History order data is integrated.
In one embodiment of this invention, the History Order data that the step-length of mobile data interception window is Dan Tian.
In one embodiment of this invention, the History Order data of the acquisition unit acquisition route include: acquisition history
The corresponding initial data of order；Initial data is pre-processed, the order for obtaining History Order data generates time and order
Amount.
In one embodiment of this invention, the pretreatment includes at least one in outlier processing or missing values processing
Kind.
In one embodiment of this invention, the training unit obtains the online data of order, is based on the online data
It trains and updates the shot and long term memory models.
It is yet another aspect of the present invention to provide a kind of logistics systems, comprising: memory can be executed for storing by processor
Instruction；Processor realizes method as described above for executing described instruction.
Another aspect of the invention provides a kind of computer-readable medium, is stored thereon with computer instruction, wherein when
When computer instruction is executed by processor, method as described above is executed.
Compared with prior art, the present invention orders the history of logistics order using the shot and long term memory models in deep learning
Forms data carries out sample extraction, the order volume that may occur in following a period of time can be predicted, so that subsequent logistics is excellent
Changing algorithm can achieve the optimal value of objective function, have the advantages that versatile, timeliness is strong, accuracy is high, intelligentized.
Detailed description of the invention
For the above objects, features and advantages of the present invention can be clearer and more comprehensible, below in conjunction with attached drawing to tool of the invention
Body embodiment elaborates, in which:
Fig. 1 is the exemplary process diagram of the order forecast method of one embodiment of the invention；
Fig. 2 is that the mobile data interception window of one embodiment of the invention obtains the schematic diagram of training sample；
Fig. 3 is the structural schematic diagram of the replicated blocks in shot and long term memory models；
Fig. 4 is the structural schematic diagram of the shot and long term memory models in one embodiment of the invention；
Fig. 5 is to be carried out in one embodiment of the invention using order volume of the housebroken shot and long term memory models to the testing time
The exemplary process diagram of prediction；
Fig. 6 is the order forecasting comparative result figure of one embodiment of the invention；
Fig. 7 is the structural block diagram of the order forecasting device of one embodiment of the invention；
Fig. 8 is the block diagram of the logistics system of one embodiment of the invention.
Specific embodiment
For the above objects, features and advantages of the present invention can be clearer and more comprehensible, below in conjunction with attached drawing to tool of the invention
Body embodiment elaborates.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, but the present invention can be with
It is different from other way described herein using other and implements, therefore the present invention is by the limit of following public specific embodiment
System.
As shown in the application and claims, unless context clearly prompts exceptional situation, " one ", "one", " one
The words such as kind " and/or "the" not refer in particular to odd number, may also comprise plural number.It is, in general, that term " includes " only prompts to wrap with "comprising"
Include clearly identify the step of and element, and these steps and element do not constitute one it is exclusive enumerate, method or apparatus
The step of may also including other or element.
Fig. 1 is the exemplary process diagram of the order forecast method of one embodiment of the invention.Refering to what is shown in Fig. 1, the order
Prediction technique the following steps are included:
Step 110, the History Order data of route are acquired.
The History Order data include corresponding to the order generation time of a certain route and order volume.Route can be by delivering
Ground and place of acceptance determine that the place of departure and place of acceptance can issue city and purpose city, such as Shanghai-Shanghai line respectively
Road, Shanghai-Suzhou route etc..Place of departure and place of acceptance are also possible to province, county-level city, area, storage code etc..For example, table one
It show by the obtained History Order data of Shanghai-Shanghai route.
Table one:
* in table one is any number for indicating order volume.The order generation time is not limited to the record form in table one, also
It may include year, month, day, hour, min, second and what day information.Since the purpose of the invention is to according to History Order
Data predict the order volume of the following some day, that is, carry out day degree order forecasting, therefore order in History Order data
Single generation time should at least generate using day as the period.In some embodiments, order generates the time also in History Order data
It can be generated using hour as the period.
In this step, the History Order data that can acquire a plurality of route simultaneously calculate every two lines in a plurality of route
The similarity of road History Order data merges the two lines road that similarity is greater than preset value, so as to increase subsequent grown
The scale of the training dataset of short-term memory model.Wherein, the phase of every two lines road History Order data in a plurality of route is calculated
It can include but is not limited to manhatton distance similarity algorithm, Euclidean distance similarity algorithm, cosine similarity like the method for degree
Algorithm or Pearson's similarity algorithm.For example, by similarity calculation the result shows that, Shanghai-Shanghai route and Shanghai-Suzhou
The similarity of route is higher, is greater than preset value, then merges the two lines road, i.e., will correspond to the History Order on the two lines road
Data are combined the History Order data as same route (such as being denoted as A route).Obviously, generated A after merging
Include all History Order data in former Shanghai-Shanghai route and Shanghai-Suzhou route in route, is acquired by A route
Obtained from the data volumes of History Order data be equal to History Order data in former Shanghai-Shanghai route and Shanghai-Suzhou route
The sum of data volume.
It is understood that can also there is the similarity between the History Order data more than two lines road to be both greater than
Preset value, then two lines road should be more than by merging.
In some embodiments, the History Order data for acquiring route may include: that acquisition History Order is corresponding original
Data；Initial data is pre-processed, the order for obtaining History Order data generates time and order volume.In some embodiments
In, pretreatment includes at least one of outlier processing or missing values processing.
Outlier processing, which can be, carries out inspection verification, excluding outlier, to keep away to the corresponding initial data of History Order
Exempting from exceptional value influences the accuracy of subsequent prediction.Missing values processing can be, and examine to the corresponding initial data of History Order
It checks pair, corresponding missing values is filled up, to improve the accuracy of order forecasting.
Step 120, History Order data are integrated according to the order generation time.
The purpose of this step is by integrating to History Order data, after being suitable for these History Order data
The training and test of continuous shot and long term memory models.For example, generating the sequence of time according to order to arrange History Order data.
It, can be in some way for data set made of being merged as the History Order data of more similar a plurality of route
The History Order data for corresponding to a plurality of route are arranged, for example, being arranged according to the sequence that order generates the time.
In some embodiments, History Order data are integrated according to the number of weeks that order generates the time.Firstly, will
Each order generates the time and arranges according to the number of weeks corresponding to it, as shown in Table 2.
Table two:
| Monday | Tuesday | Wednesday | Thursday | Friday | Saturday | Sunday |
| Day1 | Day2 | Day3 | Day4 | Day5 | Day6 | Day7 |
| Day8 | Day9 | Day10 | Day11 | Day12 | Day13 | Day14 |
| Day15 | Day16 | Day17 | Day18 | Day19 | Day20 | Day21 |
Table two illustratively lists the History Order data of the limited quantity arranged according to number of weeks.At each
It is number of days or the date that corresponding order generates the time in column corresponding to number of weeks.Again by all order generate the time according to
Number of weeks generates different data sequences, as shown in Table 3.
Table three:
| Sequence 1 | Day1 | Day8 | Day15 | …… | Day n |
| Sequence 2 | Day2 | Day9 | Day16 | …… | Day n+1 |
| Sequence 3 | Day3 | Day10 | Day17 | …… | Day n+2 |
| Sequence 4 | Day4 | Day11 | Day18 | …… | Day n+3 |
| Sequence 5 | Day5 | Day12 | Day19 | …… | Day n+4 |
| Sequence 6 | Day6 | Day13 | Day20 | …… | Day n+5 |
| Sequence 7 | Day7 | Day14 | Day21 | …… | Day n+6 |
Sequence 1 in table three separately includes the order generation time for all corresponding to Monday to Sunday into sequence 7.
According still further to the content of table three, History Order data corresponding to the time are generated to each order according to the sequence of sequence from 1 to 7
It is arranged line by line.It is assumed that each sequence includes that m order generates the time, then it is right the order in sequence 1 to be generated time institute
The History Order data arrangement answered is D1 to Dm, and the order in sequence 2 generates History Order data arrangement corresponding to the time and is
Dm+1 to D2m, and so on, it is Dm+6 to D7m that the order in sequence 7, which generates History Order data arrangement corresponding to the time,.
As shown in Table 4.
Table four:
| Sequence 1 | D1 | D2 | D3 | …… | Dm |
| Sequence 2 | Dm+1 | Dm+2 | Dm+3 | …… | D2m |
| Sequence 3 | D2m+1 | D2m+2 | D2m+3 | …… | D3m |
| Sequence 4 | D3m+1 | D3m+2 | D3m+3 | …… | D4m |
| Sequence 5 | D4m+1 | D4m+2 | D4m+3 | …… | D5m |
| Sequence 6 | D5m+1 | D5m+2 | D5m+3 | …… | D6m |
| Sequence 7 | D6m+1 | D6m+2 | D6m+3 | …… | D7m |
When shot and long term memory models are trained or are tested using these History Order data, be from table four according to
Sequence line by line arranges these History Order data, is as shown in Table 5 the number for the training of shot and long term memory models or test
According to collection:
Table five:
| D1 | D2 | D3 | …… | D15 | D16 | D17 | D18 | D19 | …… |
In the data set, D1 indicates the route in the order volume on Day1 (Monday), and D2 indicates the route in Day8 (star
Phase one) order volume, and so on, the history that sequence 2 is only until the last one History Order data Dm of sequence 1 is ordered
Forms data.
Step 130, from the N days order volumes of History Order data cutout through integrating as timing list entries, and with N
Label of+1 day order volume as the timing list entries, mobile data intercept window and obtain multiple training samples.The training
Sample includes timing list entries and label.
Fig. 2 is that the mobile data interception window of one embodiment of the invention obtains the schematic diagram of training sample.Institute is determined first
The number of days N to be intercepted.In the embodiment shown in Figure 2, N=15.Refering to what is shown in Fig. 2, the length of data cutout window is N+1=
16, which includes totally 16 days History Order data, i.e. order volume from D1 to D16.Wherein, the order volume conduct of D1 to D15
Timing list entries, i.e., timing list entries 1 as shown in Figure 2, label of the D16 as the timing list entries, i.e., such as Fig. 2
Shown in label 1.
In the embodiment shown in Figure 2, the History Order data that the step-length of mobile data interception window is Dan Tian.Such as figure
Shown in 2, after determining timing list entries 1 and label 1.Data cutout window is moved backward 1 day, to obtain timing
List entries 2 and label 2.It wherein include the order volume of D2 to D16 in timing list entries 2, label 2 is the order volume of D17.It is aobvious
So, the order volume data in timing list entries 1 and timing list entries 2 overlap.According to movement shown in Fig. 2
The method of data cutout window can obtain multiple timing list entries and label.The quantity of the timing list entries and label
It is related with the big smallest number of total data set.
It is understood that in other examples, the step-length that mobile data intercepts window can be greater than one day.In order to
All data in data set are made full use of, the step-length of mobile data interception window should be less than N, each in data set to guarantee
A data may be used to timing list entries.When the step-length is single day, prediction effect is best.
Step 140, using multiple training samples training shot and long term memory models.
Fig. 3 is the structural schematic diagram of the replicated blocks in shot and long term memory models.Shot and long term remembers (Long Short-
Term Memory, LSTM) model is a kind of special Recognition with Recurrent Neural Network (RNN) model, it is capable of the dependence of Chief Learning Officer, CLO.
Refering to what is shown in Fig. 3, containing four layers of interactive neural net layer (311-314) in the replicated blocks of the LSTM model.Across the knot
Horizontal line at the top of structure schematic diagram indicates the state of cell.It is wherein C by the cell state that previous replicated blocks exportt-1, by this
The cell state of replicated blocks output is Ct.Each neural net layer and a point multiplication operation form a door (Gate).
Wherein, neural net layer 311,312 and 313 is all Sigmoid neural net layer.Neural net layer 311 and point multiplication operation 321
It is formed and forgets gate layer.Neural net layer 312 and point multiplication operation 322 form input gate layer.Neural net layer 314 and point multiplication are transported
It calculates 323 and forms output gate layer.χ shown in Fig. 3tFor input vector,WithRespectively represent hidden layer, LSTM model
Input gate, the activation primitive parameter for forgeing door, state and out gate.L respectively indicates l layers of model, when t indicates timing input
Between t.
Fig. 4 is the structural schematic diagram of the LSTM model in one embodiment of the invention.Refering to what is shown in Fig. 4, the LSTM model can
It is together in series to form chain structure with replicated blocks as shown in Figure 3, including input layer 410, hidden layer 420 and output layer 430.Its
In each layer be made of the replicated blocks of LSTM model.Each activation primitive parameter in the LSTM model can be by following
Formula obtains:
In above-mentioned formula, W indicates that model training parameter matrix, b indicate training straggling parameter.In Fig. 4, WiIt is defeated
Enter the model training parameter matrix of layer, WfFor the model training parameter matrix of hidden layer, WoFor the model training parameter square of output layer
Battle array.
A certain number of training samples are obtained using mobile data interception window as shown in Figure 2, by these training samples
As the input of LSTM model, which is trained.For example, by taking embodiment shown in Fig. 2 as an example, each training
Sample all include have N days order volume timing list entries and the order volume with the N+1 days label.These are trained
Sample is all input in LSTM model, is trained using multiple timing list entries to LSTM model, will input sequence by timing
It arranges the N+1 days estimated order volumes to be compared with label corresponding to the timing list entries, make between the two
Error reaches minimum, then training terminates.
It is understood that corresponding to different routes, training sample is also different.Therefore, for different routes
For, the parameters in housebroken LSTM model are also different.
In some embodiments, further include the online data for obtaining order, based on online data training and update shot and long term
Memory models.By the shot and long term memory models for learning to update previous off-line training again and completing to online data, can make to grow
Short-term memory model dynamic updates, and improves the accuracy of prediction.
Step 150, the order volume of testing time is predicted using housebroken shot and long term memory models.
Fig. 5 is to be carried out in one embodiment of the invention using order volume of the housebroken shot and long term memory models to the testing time
The exemplary process diagram of prediction.Refering to what is shown in Fig. 5, the pre- flow gauge the following steps are included:
Step 510, route j N days History Order data before the testing time are collected.Here testing time is user
Want some day of prediction order volume.
Step 520, it is loaded into the housebroken shot and long term memory models parameter of corresponding line.
Step 530, N days before corresponding line History Order data are input in shot and long term memory models.
Step 540, the prediction order volume of testing time is exported.
Pre- flow gauge shown in fig. 5 can be the step of circulation, complete to route j the testing time order
After the prediction of amount, and then route j+1 can be predicted in the order volume of testing time.Wherein, for different routes
Testing time can be different.
It should be noted that when being integrated according to number of weeks to History Order data, and from the data after the integration
It extracts training sample to be trained LSTM model, which has the order volume prediction in the following week best
Effect.
Illustrate the effect of order forecast method of the invention below by a specific embodiment.The embodiment is collected
Route is Shanghai-Jiangsu route.Wherein, the order generation time of the history day degree order data in training data is from 2013
On December 31, on January 8, to 2015 in January, totally 1065 effective samples；History day degree order data in test data
It is totally 320 effective samples from December 31,1 day to 2016 January in 2016 that order, which generates the time,.To the training number of the route
Multiple training samples are obtained according to the pretreatment carried out as shown at step 120, and by step 130.With these training samples pair
LSTM model is trained, and obtains the model parameter for corresponding to the route.Further according to pre- flow gauge shown in fig. 5, to test number
It is predicted according to the data of middle every day.
Fig. 6 is the order forecasting comparative result figure of one embodiment of the invention.It is above-mentioned Shanghai-Jiangsu route shown in Fig. 6
Order forecasting comparative result figure.Horizontal axis is number of days in figure, is shared 320 days, the longitudinal axis is order volume.Order corresponding to every day
The actual value of amount is as shown in dotted line, and predicted value is by shown in solid.As it can be seen that predicted value and actual value have for the embodiment
There is the preferable goodness of fit.
It is understood that the embodiment of the present invention is to carry out by taking route as an example to order volume corresponding to certain route
Prediction.Thought according to the present invention can also acquire History Order data for different vehicles, transport power type etc., and be directed to
Order volume corresponding to these types is predicted.
Order forecast method according to the present invention can be spent on the period in day predicts the following order volume.Its advantage exists
In:
(1) versatile: order forecast method of the invention can use before any Logistic Scheduling algorithm, be next
Logistic Scheduling in a working day or even certain following a period of time provides prediction and instructs, more efficient carry out stream line optimization,
Transport capacity dispatching and haul-cycle time optimization.
(2) timeliness is strong, and accuracy is high: the present invention uses intelligent algorithm, realize intelligently to historical data into
Row is analyzed and is predicted the following order volume, and prediction timeliness is strong, accuracy is high, can be realized the high frequency prediction of day degree, and
And there is very high prediction accuracy.
(3) intelligent: the present invention can intelligently realize self-teaching, optimization by intelligent algorithm, and can be certainly
It moves according to the logistics conditions such as every stream line, vehicle, transport power type, adaptively self-optimization, there is very high degree of intelligence.
Fig. 7 is the structural block diagram of the order forecasting device of one embodiment of the invention.Refering to what is shown in Fig. 7, the device includes adopting
Collect unit 710, integral unit 720, training sample acquiring unit 730, training unit 740 and predicting unit 750.Wherein, it acquires
Unit 710 is used to acquire the History Order data of route, which includes that order generates time and order volume.Specifically
Ground method can refer to step 110 shown in FIG. 1 and its description.
In some embodiments, the History Order data of acquisition unit acquisition route may include: to obtain History Order pair
The initial data answered；Initial data is pre-processed, the order for obtaining History Order data generates time and order volume.One
In a little embodiments, pretreatment may include at least one of outlier processing or missing values processing.
Outlier processing, which can be, carries out inspection verification, excluding outlier, to keep away to the corresponding initial data of History Order
Exempting from exceptional value influences the accuracy of subsequent prediction.Missing values processing can be, and examine to the corresponding initial data of History Order
It checks pair, corresponding missing values is filled up, to improve the accuracy of order forecasting.
Integral unit 720 is used to integrate History Order data according to the order generation time, specific integration side
Method can refer to step 120 shown in FIG. 1 and its description.
Training sample acquiring unit 730 is used for from the N days order volumes of History Order data cutout through integrating as timing
List entries, and using the N+1 days order volumes as the label of the timing list entries, it is multiple that mobile data intercepts window acquisition
Training sample, the training sample include timing list entries and label.Specific training sample acquisition methods can refer to Fig. 1 institute
The step 130 and its description shown.Training unit 740 is used for using multiple as acquired in training sample acquiring unit 730
Training sample trains LSTM model.Specific training method can refer to step 140 shown in FIG. 1 and its description.
In some embodiments, training unit obtains the online data of order, described in online data training and updating
Shot and long term memory models.It, can by the shot and long term memory models for learning to update previous off-line training again and completing to online data
So that shot and long term memory models dynamic updates, the accuracy of prediction is improved.
Predicting unit 750 is for predicting the order volume of testing time using housebroken LSTM model.Specifically
Prediction technique can refer to step 150 shown in FIG. 1 and its description.
Technical solution of the present invention also proposed a kind of logistics system, including memory and processor.Wherein, memory is used
In the instruction that storage can be executed by processor；Processor is for executing the instruction to realize that order shown in the embodiment of the present invention is pre-
Survey method.Fig. 8 is the block diagram of the logistics system of one embodiment of the invention.Refering to what is shown in Fig. 8, the logistics system may include inside
Communication bus 801, processor 802, read-only memory (ROM) 803, random access memory (RAM) 804, communication port 805,
And hard disk 806.The data communication inside logistics system between various components may be implemented in internal communication bus 801.Processor
802 can be judged and be issued prompt.In some embodiments, processor 802 can be made of one or more processors.
Communication port 805 may be implemented logistics system and external module, such as client etc., between carry out data communication.
In some embodiments, logistics system may be coupled to network by communication port 805, and sends and connect from network
By information and data.The connection can be wired connection, be wirelessly connected, can be realized data transmission and/or it is received it is any its
He communicates to connect, and/or any combination of these connections.Wired connection may include such as cable, optical cable, telephone wire or its
Any combination.Wireless connection may include such as bluetoothTMLink, Wi-FiTMLink, WiMaxTMLink, WLAN link, ZigBee chain
Road, mobile network's link (for example, 3G, 4G, 5G etc.) etc. or combinations thereof.In some embodiments, communication port 805 can be packet
Include Standardization Communication port, RS232, RS485 etc..
Memory in logistics system may include various forms of program storage units and data storage element, such as
Hard disk 806, read-only memory 803 and random access memory 804, can store computer disposal and/or communication use it is each
Possible program instruction performed by kind data file and processor 802.Processor 802 in logistics system executes these
Instruct the major part to realize the method for prediction order of the invention.The result of the processing of processor 802 passes through communication port 805
It is transmitted to client, and can be shown on a user interface.
Order forecast method of the invention can be stored in computer-readable medium in the form of computer program code,
The computer program code is executed by processor to realize that this method wraps function to be achieved.
The some aspects of order forecast method of the invention can completely by hardware execute, can completely by software (including
Firmware, resident software, microcode etc.) it executes, can also be executed by combination of hardware.Hardware above or software are referred to alternatively as
" data block ", " module ", " engine ", " unit ", " component " or " system ".Processor can be one or more dedicated integrated electricity
Road (ASIC), digital signal processor (DSP), digital signal processing device (DAPD), programmable logic device (PLD), scene
Programmable gate array (FPGA), processor, controller, microcontroller, microprocessor or a combination thereof.In addition, the application's is each
Aspect may show as the computer product being located in one or more computer-readable mediums, which includes computer-readable
Program coding.For example, computer-readable medium may include, but be not limited to, magnetic storage device is (for example, hard disk, floppy disk, magnetic
Band ...), CD (for example, compact disk CD, digital versatile disc DVD ...), smart card and flash memory device (for example, card,
Stick, key drive ...).
Computer-readable medium may include the propagation data signal containing computer program code in one, such as in base
Take or as carrier wave a part.There are many transmitting signal possibility form of expression, including electromagnetic form, light form etc.,
Or suitable combining form.Computer-readable medium can be any computer-readable in addition to computer readable storage medium
Medium, the medium can realize communication, propagation or transmission for making by being connected to an instruction execution system, device or equipment
Program.Program coding on computer-readable medium can be propagated by any suitable medium, including nothing
The combination of line electricity, cable, fiber optic cables, radiofrequency signal or similar mediums or any of above medium.
In addition, except clearly stating in non-claimed, the sequence of herein described processing element and sequence, digital alphabet
Using or other titles use, be not intended to limit the sequence of the application process and method.Although by each in above-mentioned disclosure
Kind of example discuss it is some it is now recognized that useful inventive embodiments, but it is to be understood that, such details only plays explanation
Purpose, appended claims are not limited in the embodiment disclosed, on the contrary, claim is intended to cover and all meets the application
The amendment and equivalent combinations of embodiment spirit and scope.For example, although system component described above can be set by hardware
It is standby to realize, but can also be only achieved by the solution of software, such as pacify on existing server or mobile device
Fill described system.
Similarly, it is noted that in order to simplify herein disclosed statement, to help real to one or more application
Apply the understanding of example, above in the description of the embodiment of the present application, sometimes by various features merger to one embodiment, attached drawing or
In descriptions thereof.But this disclosure method is not meant to mention in aspect ratio claim required for the application object
And feature it is more.In fact, the feature of embodiment will be less than whole features of the single embodiment of above-mentioned disclosure.
The number of description ingredient, number of attributes is used in some embodiments, it should be appreciated that such to be used for embodiment
The number of description has used qualifier " about ", " approximation " or " generally " to modify in some instances.Unless in addition saying
It is bright, " about ", " approximation " or " generally " show the variation that the number allows to have ± 20%.Correspondingly, in some embodiments
In, numerical parameter used in description and claims is approximation, approximation feature according to needed for separate embodiment
It can change.In some embodiments, numerical parameter is considered as defined significant digit and using the reservation of general digit
Method.Although the Numerical Range and parameter in some embodiments of the application for confirming its range range are approximation, specific real
It applies in example, being set in for such numerical value is reported as precisely as possible in feasible region.
Although the present invention is described with reference to current specific embodiment, those of ordinary skill in the art
It should be appreciated that above embodiment is intended merely to illustrate the present invention, can also make in the case where no disengaging spirit of that invention
Various equivalent change or replacement out, therefore, as long as to the variation of above-described embodiment, change in spirit of the invention
Type will all be fallen in the range of following claims.
Claims (10)
1. a kind of order forecast method, this method comprises:
The History Order data of route are acquired, the History Order data include that order generates time and order volume；
The History Order data are integrated according to the order generation time；
From the N days order volumes of History Order data cutout through integrating as timing list entries, and with the N+1 days order volumes
As the label of the timing list entries, mobile data intercepts window and obtains multiple training samples, when the training sample includes
Sequence list entries and label；
Use the multiple training sample training shot and long term memory models；
The order volume of testing time is predicted using housebroken shot and long term memory models.
2. order forecast method as described in claim 1, which is characterized in that acquire the History Order data of a plurality of route, count
The similarity of every two lines road History Order data in a plurality of route is calculated, the two lines road that similarity is greater than preset value is merged.
3. order forecast method as claimed in claim 2, which is characterized in that using manhatton distance similarity algorithm, European
Distance conformability degree algorithm, cosine similarity algorithm or Pearson's similarity algorithm calculate every two lines road history in a plurality of route and order
The similarity of forms data.
4. order forecast method as described in claim 1, which is characterized in that generate the number of weeks pair of time according to the order
The History Order data are integrated.
5. order forecast method as described in claim 1, which is characterized in that the step-length that mobile data intercepts window is Dan Tian's
History Order data.
6. order forecast method as described in claim 1, which is characterized in that the History Order data for acquiring route include:
Obtain the corresponding initial data of History Order；
Initial data is pre-processed, the order for obtaining History Order data generates time and order volume.
7. order forecast method as claimed in claim 6, which is characterized in that the pretreatment includes outlier processing or missing
At least one of value processing.
8. a kind of order forecasting device, the device include:
Acquisition unit, acquires the History Order data of route, and the History Order data include that order generates time and order volume；
Integral unit integrates the History Order data according to the order generation time；
Training sample acquiring unit, from the N days order volumes of History Order data cutout through integrating as timing list entries, and
Using the N+1 days order volumes as the label of the timing list entries, mobile data intercepted window and obtains multiple training samples, institute
Stating training sample includes timing list entries and label；
Training unit uses the multiple training sample training shot and long term memory models；
Predicting unit predicts the order volume of testing time using housebroken shot and long term memory models.
9. a kind of logistics system, comprising:
Memory, for storing the instruction that can be executed by processor；
Processor, for executing described instruction to realize the method according to claim 1 to 7.
10. a kind of computer-readable medium, is stored thereon with computer instruction, wherein when computer instruction is executed by processor
When, execute the method according to claim 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910382417.1A CN110097320A (en) | 2019-05-09 | 2019-05-09 | Order forecast method and device, logistics system and computer-readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910382417.1A CN110097320A (en) | 2019-05-09 | 2019-05-09 | Order forecast method and device, logistics system and computer-readable medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110097320A true CN110097320A (en) | 2019-08-06 |
Family
ID=67447483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910382417.1A Pending CN110097320A (en) | 2019-05-09 | 2019-05-09 | Order forecast method and device, logistics system and computer-readable medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110097320A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111191949A (en) * | 2020-01-03 | 2020-05-22 | 北京三快在线科技有限公司 | Method and device for identifying abnormal driving behaviors of networked vehicles and electronic equipment |
| CN111242747A (en) * | 2020-01-21 | 2020-06-05 | 北京爱数智慧科技有限公司 | Order quotation method and device |
| CN112308514A (en) * | 2020-10-30 | 2021-02-02 | 康键信息技术（深圳）有限公司 | Logistics order automatic issuing method, device, equipment and storage medium |
| CN112837001A (en) * | 2019-11-22 | 2021-05-25 | 顺丰科技有限公司 | Logistics network planning method and device and computer readable storage medium |
| CN112948763A (en) * | 2019-12-11 | 2021-06-11 | 顺丰科技有限公司 | Method and device for predicting quantity of component, electronic equipment and storage medium |
| CN114037130A (en) * | 2021-10-28 | 2022-02-11 | 重庆川南环保科技有限公司 | Household appliance maintenance order prediction method and device and terminal equipment |
| CN116720202A (en) * | 2023-05-19 | 2023-09-08 | 国网物资有限公司 | Service information detection method, device, electronic equipment and computer readable medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103310286A (en) * | 2013-06-25 | 2013-09-18 | 浙江大学 | Product order prediction method and device with time series characteristics |
| CN108564326A (en) * | 2018-04-19 | 2018-09-21 | 安吉汽车物流股份有限公司 | Prediction technique and device, computer-readable medium, the logistics system of order |
| CN108960537A (en) * | 2018-08-17 | 2018-12-07 | 安吉汽车物流股份有限公司 | The prediction technique and device of logistics order, readable medium |
-
2019
- 2019-05-09 CN CN201910382417.1A patent/CN110097320A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103310286A (en) * | 2013-06-25 | 2013-09-18 | 浙江大学 | Product order prediction method and device with time series characteristics |
| CN108564326A (en) * | 2018-04-19 | 2018-09-21 | 安吉汽车物流股份有限公司 | Prediction technique and device, computer-readable medium, the logistics system of order |
| CN108960537A (en) * | 2018-08-17 | 2018-12-07 | 安吉汽车物流股份有限公司 | The prediction technique and device of logistics order, readable medium |
Non-Patent Citations (1)
| Title |
|---|
| 黄冬梅等: "海洋生物酶发酵过程软测量与优化控制方法研究", 上海交通大学出版社 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112837001A (en) * | 2019-11-22 | 2021-05-25 | 顺丰科技有限公司 | Logistics network planning method and device and computer readable storage medium |
| CN112948763A (en) * | 2019-12-11 | 2021-06-11 | 顺丰科技有限公司 | Method and device for predicting quantity of component, electronic equipment and storage medium |
| CN112948763B (en) * | 2019-12-11 | 2024-04-09 | 顺丰科技有限公司 | Piece quantity prediction method and device, electronic equipment and storage medium |
| CN111191949A (en) * | 2020-01-03 | 2020-05-22 | 北京三快在线科技有限公司 | Method and device for identifying abnormal driving behaviors of networked vehicles and electronic equipment |
| CN111191949B (en) * | 2020-01-03 | 2022-06-17 | 北京三快在线科技有限公司 | Method and device for identifying abnormal driving behaviors of networked vehicles and electronic equipment |
| CN111242747A (en) * | 2020-01-21 | 2020-06-05 | 北京爱数智慧科技有限公司 | Order quotation method and device |
| CN112308514A (en) * | 2020-10-30 | 2021-02-02 | 康键信息技术（深圳）有限公司 | Logistics order automatic issuing method, device, equipment and storage medium |
| CN114037130A (en) * | 2021-10-28 | 2022-02-11 | 重庆川南环保科技有限公司 | Household appliance maintenance order prediction method and device and terminal equipment |
| CN116720202A (en) * | 2023-05-19 | 2023-09-08 | 国网物资有限公司 | Service information detection method, device, electronic equipment and computer readable medium |
| CN116720202B (en) * | 2023-05-19 | 2024-03-29 | 国网物资有限公司 | Service information detection method, device, electronic equipment and computer readable medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110097320A (en) | Order forecast method and device, logistics system and computer-readable medium | |
| CN110110932A (en) | Order forecast method and device, logistics system and computer-readable medium | |
| CN110110931A (en) | Order forecast method and device, logistics system and computer-readable medium | |
| Gasparin et al. | Deep learning for time series forecasting: The electric load case | |
| US11423336B2 (en) | Method and system for model integration in ensemble learning | |
| CN111415048B (en) | Vehicle path planning method based on reinforcement learning | |
| Gilik et al. | Air quality prediction using CNN+ LSTM-based hybrid deep learning architecture | |
| CN110782093B (en) | PM fusing SSAE deep feature learning and LSTM2.5Hourly concentration prediction method and system | |
| CN108694673A (en) | A kind of processing method, device and the processing equipment of insurance business risk profile | |
| CN107092974A (en) | Dispense pressure prediction method and device | |
| CN111063194A (en) | Traffic flow prediction method | |
| CN103747523A (en) | User position predicating system and method based on wireless network | |
| CN106022549A (en) | Short term load predication method based on neural network and thinking evolutionary search | |
| CN107330464A (en) | Data processing method and device | |
| CN109543924A (en) | Goods amount prediction technique, device and computer equipment | |
| CN111079975A (en) | Traffic data prediction method and device and vehicle control method | |
| CN112597217B (en) | Intelligent decision platform driven by historical decision data and implementation method thereof | |
| Wang et al. | Forecasting hotel reservations with long short-term memory-based recurrent neural networks | |
| Chen et al. | Data mining-based dispatching system for solving the local pickup and delivery problem | |
| CN113919558A (en) | Product sales prediction method and device | |
| CN111507499B (en) | Method, device and system for constructing model for prediction and testing method | |
| CN117035207A (en) | Method for planning path of securicar, method and device for training path prediction model | |
| CN108596390B (en) | Method for solving vehicle path problem | |
| CN111325401B (en) | Method and device for training path planning model and computer system | |
| CN116662815B (en) | Training method of time prediction model and related equipment |
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: 20190806 |
|
| RJ01 | Rejection of invention patent application after publication |