CN108564326A - Prediction technique and device, computer-readable medium, the logistics system of order - Google Patents
Prediction technique and device, computer-readable medium, the logistics system of order Download PDFInfo
- Publication number
- CN108564326A CN108564326A CN201810357890.XA CN201810357890A CN108564326A CN 108564326 A CN108564326 A CN 108564326A CN 201810357890 A CN201810357890 A CN 201810357890A CN 108564326 A CN108564326 A CN 108564326A
- Authority
- CN
- China
- Prior art keywords
- order
- neural network
- sequence
- network model
- prediction
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
-
- 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"
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of prediction technique and device, computer-readable medium, logistics system of order, the prediction technique of the order include：Obtain the corresponding time series of History Order；Based on the time series, characteristic sequence is extracted；Based on the characteristic sequence, 2D signal characteristic pattern is generated；Based on the 2D signal characteristic pattern, neural network model is built, and order forecasting is carried out according to constructed neural network model.Using said program, the accuracy of order forecasting can be improved.
Description
Technical field
The present embodiments relate to the prediction techniques and device of logistics field more particularly to a kind of order, computer-readable
Medium, logistics system.
Background technology
It since the order forecasting of complete vehicle logistics can allow dispatcher to be ready in advance, provides for a rainy day, so that fortune
The scheduling of defeated resource is more reasonable, therefore accurately the order forecasting of complete vehicle logistics has very important work for complete vehicle logistics
With.
Time series can describe the centrifugal pump that variable changes over time, such as temperature, humidity, stock price, automobile pin
Measure the variables such as tendency.Time series forecasting refer to according to time series current time point t and passing time point (t-1),
(t-2) ..., the value of (t-n-1) determines some unknown time point (t+1) or some unknown period (t+1, t+n)
Value.In general, a time series forecasting value other than dependent on the passing time, can may also depend on one or more
Other time serieses.Therefore, a time series forecasting problem refers to the regression problem in higher dimensional space.In higher dimensional space
In, predicted value is the nonlinearity function of its past value and other relevant time serieses.The order forecasting of complete vehicle logistics is asked
Topic is essentially considered as a time series forecasting problem.Complete vehicle logistics order forecasting problem is due to substantially can quilt
It is regarded as a time series forecasting problem, so can be solved by time series predicting model, such as traditional autoregression moves
Dynamic averaging model (Auto Regressive Moving Average, ARMA), BP network model, Recognition with Recurrent Neural Network
(Recurrent Neural Network, RNN), variant length Memory Neural Networks (Long Short-Term Memory,
LSTM) etc..These models are merged into one-dimensional data when solving the problems, such as complete vehicle logistics order forecasting, by logistics order data and are made
For input, the temporal correlation in data is then extracted.
The prediction technique of existing complete vehicle logistics order is having multiple time serieses or multiple characteristic sequences as input
In the case of, the method by being simply merged into a sequence is carried out an one-dimensional sequence of synthesis as mode input
Study and prediction.
The prediction technique accuracy of above-mentioned complete vehicle logistics order is poor.
Invention content
The technical issues of embodiment of the present invention solves is how to improve the accuracy of order forecasting.
In order to solve the above technical problems, the embodiment of the present invention provides a kind of prediction technique of order, the method includes：It obtains
Take the corresponding time series of History Order；Based on the time series, characteristic sequence is extracted；Based on the characteristic sequence, generate
2D signal characteristic pattern；Based on the 2D signal characteristic pattern, neural network model is built, and according to constructed neural network
Model carries out order forecasting.
Optionally, the corresponding time series of the acquisition order includes：Obtain the corresponding initial data of History Order；To original
Beginning data are pre-processed, and the corresponding time series of order is obtained.
Optionally, the pretreatment includes following at least one：Outlier processing, missing values processing.
Optionally, the extraction characteristic sequence includes：Characteristic sequence is extracted based on Wavelet Transformation Algorithm.
Optionally, the generation 2D signal characteristic pattern includes：Each characteristic sequence is divided into the sequence that multiple length are n
Column-slice section, wherein n are positive integer；The corresponding sequence fragment of different characteristic sequence is replicated line by line, generates m row tracts
Section, it is adjacent two-by-two in the ranks that the m rows sequence fragment meets different characteristic sequences, and wherein m is positive integer；Based on the m rows sequence
Column-slice section generates the 2D signal characteristic pattern of m*n.
Optionally, it is described by each characteristic sequence be divided into multiple length be n sequence fragment include：It is grasped based on displacement
Make, each characteristic sequence is divided into the sequence fragment that multiple length are n.
Optionally, the structure neural network model includes：Build neural network model；Based on the 2D signal feature
Figure, the training neural network model, obtains the neural network model parameter.
Optionally, after building neural network model, the prediction technique of the order further includes：Obtain the online of order
Data；Based on online data training and update the neural network model.
Optionally, the neural network is：Convolutional neural networks.
The embodiment of the present invention provides a kind of prediction meanss of order, including：First acquisition unit is suitable for obtaining History Order
Corresponding time series；Extraction unit is suitable for being based on the time series, extracts characteristic sequence；Generation unit is suitable for being based on institute
Characteristic sequence is stated, 2D signal characteristic pattern is generated；Construction unit is suitable for being based on the 2D signal characteristic pattern, builds nerve net
Network model, and order forecasting is carried out according to constructed neural network model.
Optionally, the first acquisition unit includes：First obtains subelement, is suitable for obtaining History Order corresponding original
Data；Second obtains subelement, suitable for being pre-processed to initial data, obtains the corresponding time series of order.
Optionally, the pretreatment includes following at least one：Outlier processing, missing values processing.
Optionally, the extraction unit is suitable for extracting characteristic sequence based on Wavelet Transformation Algorithm.
Optionally, the generation unit includes：Divide subelement, is suitable for each characteristic sequence is divided into multiple length
The sequence fragment of n, wherein n are positive integer；Subelement is replicated, is suitable for carrying out the corresponding sequence fragment of different characteristic sequence line by line
It replicates, generates m row sequence fragments, it is adjacent two-by-two in the ranks that the m rows sequence fragment meets different characteristic sequences, and wherein m is just
Integer；Subelement is generated, is suitable for being based on the m rows sequence fragment, generates the 2D signal characteristic pattern of m*n.
Optionally, the segmentation subelement is suitable for being based on shifting function, and each characteristic sequence, which is divided into multiple length, is
The sequence fragment of n.
Optionally, the construction unit includes：Subelement is built, is suitable for being based on the neural network model parameter, structure
Neural network model；Training subelement, is suitable for being based on the 2D signal characteristic pattern, and the training neural network model obtains
The neural network model parameter；Subelement is predicted, suitable for carrying out order forecasting according to the neural network model trained.
Optionally, the prediction meanss of the order further include：Second acquisition unit is suitable for obtaining the online data of order；
Updating unit, suitable for being based on the online data training and updating the neural network model.
Optionally, the neural network is：Convolutional neural networks.
The embodiment of the present invention provides a kind of computer readable storage medium, is stored thereon with computer instruction, the calculating
The step of any of the above-described kind of the method being executed when machine instruction operation.
The embodiment of the present invention provides a kind of logistics system, including memory and processor, and being stored on the memory can
The computer instruction run on the processor, perform claim is any of the above-described when the processor runs the computer instruction
The step of kind the method.
Compared with prior art, the technical solution of the embodiment of the present invention has the advantages that：
The embodiment of the present invention generates 2D signal characteristic pattern, is then based on the two dimension by being based on the characteristic sequence
Signal characteristic figure builds neural network model, and carries out order forecasting according to constructed neural network model.Due to will be one-dimensional
Correlated series merge into two dimensional character figure, and utilize the temporally adjacent point and each phase of each sequence of neural network learning
Relevance between flanking sequence point in the same time, farthest learning signal characteristic pattern contain feature, can effectively carry
The accuracy of high order forecasting.
Further, by carrying out outlier processing to the corresponding initial data of History Order, can with rejecting abnormalities value, with
Exceptional value is avoided to influence the accuracy of subsequent prediction.
Further, by carrying out missing values processing to the corresponding initial data of History Order, to improve order forecasting
Accuracy.
Description of the drawings
Fig. 1 is a kind of flow chart of the prediction technique of order provided in an embodiment of the present invention；
Fig. 2 is the flow chart of the prediction technique of another order provided in an embodiment of the present invention；
Fig. 3 is the flow chart of the prediction technique of another order provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of the prediction meanss of order provided in an embodiment of the present invention.
Specific implementation mode
Relevance of the existing method between multiple sequences or multiple features can not learn well, cause to input number
It cannot be arrived very well by machine learning model study according to the feature contained.Therefore, the accuracy of order forecasting is poor.
The embodiment of the present invention generates 2D signal characteristic pattern, is then based on the two dimension by being based on the characteristic sequence
Signal characteristic figure builds neural network model, and carries out order forecasting according to constructed neural network model.Due to will be one-dimensional
Correlated series merge into two dimensional character figure, and utilize the temporally adjacent point and each phase of each sequence of neural network learning
Relevance between flanking sequence point in the same time, farthest learning signal characteristic pattern contain feature, can effectively carry
Rise the accuracy of order forecasting.
It is understandable to enable above-mentioned purpose, feature and the advantageous effect of the present invention to become apparent, below in conjunction with the accompanying drawings to this
The specific embodiment of invention is described in detail.
Referring to Fig. 1, an embodiment of the present invention provides a kind of prediction techniques of order, may include steps of：
Step S101 obtains the corresponding time series of History Order.
In specific implementation, in order to predict order, it is necessary first to obtain the corresponding time series of History Order, i.e.,
The related data of History Order is then based on the Future Data of the related data prediction order of History Order, i.e. higher dimensional space
Regression problem.In higher dimensional space, predicted value is the nonlinearity function of its past value and other relevant time serieses.
In specific implementation, possibly desired time series can not be directly acquired, it is corresponding whole that History Order can only be obtained
The initial data such as vehicle logistics order time series data, inventory's sequence data.Therefore the corresponding original number of History Order can be obtained
According to then being pre-processed to initial data by certain algorithm and logic, for example, data cleansing, it is corresponding to obtain order
Time series.
In an embodiment of the present invention, described pre-process includes：Outlier processing.For example, to the corresponding original of History Order
Beginning data carry out inspection verification, and rejecting abnormalities value influences the accuracy of subsequent prediction to avoid exceptional value.
In an alternative embodiment of the invention, described pre-process includes：Missing values processing.For example, corresponding to History Order
Initial data carries out inspection verification, fills up corresponding missing values, to improve the accuracy of order forecasting.
In still another embodiment of the process, missing values processing and outlier processing exist simultaneously.
It is understood that other pretreatments can also be carried out, details are not described herein again.
Step S102 is based on the time series, extracts characteristic sequence.
In specific implementation, Wavelet Transformation Algorithm can be based on and extracts the corresponding characteristic sequence of the time series.
In specific implementation, for pretreated time series, all features can be extracted first, for example, extraction M
Then a different feature carries out correlation analysis judgement, N number of maximally related characteristic sequence is chosen from M characteristic sequence,
Middle N, M are positive integer, and N≤M.
Step S103 is based on the characteristic sequence, generates 2D signal characteristic pattern.
In specific implementation, each characteristic sequence can be divided into the sequence fragment that multiple length are n, wherein n is just
Integer；Then the corresponding sequence fragment of different characteristic sequence is replicated line by line, generates m row sequence fragments, the m rows sequence
It is adjacent two-by-two in the ranks that segment meets different characteristic sequences, and wherein m is positive integer, i.e., is corresponded to each characteristic sequence identical
The sequence fragment of index is replicated, generation m row sequence fragments, and in the m rows sequence fragment, any two characteristic sequence corresponds to
The sequence fragment of same index is located at two adjacent rows, and wherein m is positive integer, is that characteristic sequence corresponds to same index per a line
Sequence fragment；It is finally based on the m rows sequence fragment, generates the 2D signal characteristic pattern of m*n.
In an embodiment of the present invention, it is based on shifting function, each characteristic sequence is divided into the sequence that multiple length are n
Segment.
In specific implementation, the shifting function can be every time one bit of displacement, or displacement two every time
The above bit can flexibly be selected according to actual conditions, and the embodiment of the present invention is not limited.
For example, extracting 5 characteristic sequences, respectively a, b, c, d, e, and the length of each characteristic sequence based on step S102
For L, i.e., each characteristic sequence includes L bit information.First by shifting slice segmentation, by 5 characteristic sequences that length is L
It is divided into the sequence fragment that L-n length is n respectively, respectively：Sequence fragment 1 corresponds to the 1 to the n-th ratio of characteristic sequence
Special, sequence fragment 2 correspond to characteristic sequence the 2 to the (n+1)th bit ..., sequence fragment L-n correspond to the L- of characteristic sequence
N to L bits；Then it generates and meets different characteristic sequences adjacent m row sequence fragments two-by-two in the ranks, for 5 features
Sequence a, b, c, d, e are generated and are met in the ranks adjacent m row sequence fragments two-by-two and be：Abcdeacebda, m=11 meet arbitrary
Two different characteristic sequences are adjacent in the ranks.It is finally based on the m rows sequence fragment, the 2D signal feature of m*n can be generated
Figure.
Step S104 is based on the 2D signal characteristic pattern, builds neural network model, and according to constructed nerve net
Network model carries out order forecasting.
In specific implementation, it can be based on the 2D signal characteristic pattern, calculate the neural network model parameter；Then
Based on the neural network model parameter, neural network model is built.
In specific implementation, convolutional neural networks may be used and build the neural network model, convolutional neural networks
(Convolutional Neural Network, CNN) is a kind of feedforward neural network, its artificial neuron can respond one
Surrounding cells in partial coverage have outstanding performance for large-scale image procossing.It includes convolutional layer (alternating
Convolutional layer) and pond layer (pooling layer).It has shift invariant, weights shared etc. good special
Property, abstract characteristics can be extracted well.
In specific implementation, the performance plot can be divided into and trains, verifies, tests several parts, and input the nerve
Network model completes the process of training and adjusting parameter.The neural network model that training is completed can be used as the mould of order forecasting
Type.
In specific implementation, over time, new logistics order data can continuously be input to the god
Through in network system, therefore it is also based on real-time online data, training simultaneously updates the neural network model, by online
The convolutional neural networks model for learning to update previous off-line training again and completing of data, makes it that can also grow with each passing hour.
Using said program, since one-dimensional correlated series are merged into two dimensional character figure, and neural network learning is utilized
The temporally adjacent point of each sequence and each mutually relevance between flanking sequence point in the same time, farthest study letter
Number characteristic pattern contains feature, can effectively promote the accuracy of order forecasting.
To make those skilled in the art be better understood from and implementing the present invention, the embodiment of the present invention gives using the present invention
The accuracy comparing result of the order forecasting of scheme and existing scheme that embodiment provides, as shown in the chart.
Table 1
Referring to table 1, based on the complete vehicle logistics order data between 2013-2016 4 years, verification is provided in an embodiment of the present invention
The accuracy of the order forecasting of scheme and existing scheme.Tri- annual datas of 2013-2015 generate neural network mould as training data
Type, is then based on generated neural network, and the order data of prediction 2016 is simultaneously carried out with true order data in 2016
It compares, provides error rate, specific formula for calculation is：Error rate=(actual value-predicted value)/actual value.
As can be seen from Table 1, using scheme provided in an embodiment of the present invention, the error of order forecasting can be effectively reduced
Rate, to provide the accuracy of higher order forecasting.
To make those skilled in the art be better understood from and implementing the present invention, an embodiment of the present invention provides another orders
Prediction technique, as shown in Figure 2.
Referring to Fig. 2, the prediction technique of the order may include steps of：
Step S201 obtains off-line indent data.
Step S202, pre-processes order data.
Step S203 is based on the order data, extracts characteristic sequence.
Step S204 builds 2D signal characteristic pattern, and is divided into training set, verification collection and test set.
Step S205 builds convolutional neural networks.
In specific implementation, the convolutional neural networks are the convolutional neural networks model for carrying out order forecasting.
Step S206 is based on the 2D signal characteristic pattern, the training convolutional neural networks.
In specific implementation, the 2D signal characteristic pattern, the training convolutional neural networks, to correct can be based on
State the parameter of convolutional neural networks.
Step S207, judges whether the convolutional neural networks meet stop condition, when the convolutional neural networks meet
When the stop condition, step S208 is executed, it is no to then follow the steps S206.
In specific implementation, the stop condition can be：Collected based on the verification, the convolutional neural networks of calculating
Error no longer decline, i.e. convolutional neural networks convergence.
Step S208, judges whether the error of the convolutional neural networks meets preset thresholding, when the convolutional Neural
When the error of network meets preset thresholding, step S209 is executed, it is no to then follow the steps S210.
Step S209 preserves the convolutional neural networks model for order forecasting.
Step S210 adjusts the convolutional neural networks structure and parameter, repeats step S205.
To make those skilled in the art be better understood from and implementing the present invention, an embodiment of the present invention provides another orders
Prediction technique, as shown in Figure 3.
Step S301 obtains online order data.
Step S302, pre-processes order data.
Step S303 is based on the order data, extracts characteristic sequence.
Step S304 builds 2D signal characteristic pattern.
Step S305 is based on on-line study, updates the convolutional neural networks model.
Step S306 is based on the convolutional neural networks model, predicts order.
Step S307 exports prediction result.
To make those skilled in the art be better understood from and implementing the present invention, the embodiment of the present invention additionally provides one kind can
Realize the prediction meanss of the prediction technique of the above order, as shown in Figure 4.
Referring to Fig. 4, the prediction meanss of the order include：First acquisition unit 41, extraction unit 42,43 and of generation unit
Construction unit 44, wherein：
The first acquisition unit 41 is suitable for obtaining the corresponding time series of History Order.
The extraction unit 42 is suitable for being based on the time series, extracts characteristic sequence.
The generation unit 43 is suitable for being based on the characteristic sequence, generates 2D signal characteristic pattern.
The construction unit 44 is suitable for being based on the 2D signal characteristic pattern, builds neural network model, and according to institute's structure
The neural network model built carries out order forecasting.
In specific implementation, the first acquisition unit 41 may include：First, which obtains subelement 411 and second, obtains son
Unit 412, wherein：
Described first obtains subelement 411, is suitable for obtaining the corresponding initial data of History Order.
Described second obtains subelement 412, suitable for being pre-processed to initial data, obtains the corresponding time sequence of order
Row.
In an embodiment of the present invention, the pretreatment includes following at least one：Outlier processing, missing values processing.
In an embodiment of the present invention, the extraction unit 42 is suitable for extracting characteristic sequence based on Wavelet Transformation Algorithm.
In specific implementation, the generation unit 43 includes：Divide subelement 431, replicate subelement 432 and generate son and is single
Member 433, wherein：
The segmentation subelement 431, suitable for each characteristic sequence is divided into the sequence fragment that multiple length are n, wherein n
For positive integer.
The duplication subelement 432 generates m suitable for being replicated the corresponding sequence fragment of different characteristic sequence line by line
Row sequence fragment, it is adjacent two-by-two in the ranks that the m rows sequence fragment meets different characteristic sequences.
The generation subelement 433 is suitable for being based on the m rows sequence fragment, generates the 2D signal characteristic pattern of m*n.
In an embodiment of the present invention, the segmentation subelement 431 is suitable for being based on shifting function, by each characteristic sequence
It is divided into the sequence fragment that multiple length are n.
In specific implementation, the construction unit 44 includes：It is single to build subelement 441, training subelement 442 and prediction
Member 443, wherein：
The structure subelement 441 is suitable for being based on the neural network model parameter, builds neural network model.
The trained subelement 442, is suitable for being based on the 2D signal characteristic pattern, and the training neural network model obtains
Take the neural network model parameter.
The prediction subelement 443, suitable for carrying out order forecasting according to the neural network model trained.
In specific implementation, the prediction meanss 40 of the order can also include：Second acquisition unit (not shown) and more
New unit (not shown), wherein：
The second acquisition unit is suitable for obtaining the online data of order.
The updating unit, suitable for being based on the online data training and updating the neural network model.
In an embodiment of the present invention, the neural network is：Convolutional neural networks.
In specific implementation, the workflow of the prediction meanss 40 of the order and principle can refer in above-described embodiment
Description in the method for offer, details are not described herein again.
The embodiment of the present invention provides a kind of computer readable storage medium, and computer readable storage medium is non-volatile deposits
Storage media or non-transitory storage media, are stored thereon with computer instruction, and the computer instruction executes any of the above-described when running
Step corresponding to kind the method, details are not described herein again.
The embodiment of the present invention provides a kind of logistics system, including memory and processor, and energy is stored on the memory
Enough computer instructions run on the processor, the processor execute any of the above-described kind when running the computer instruction
Step corresponding to the method, details are not described herein again.
One of ordinary skill in the art will appreciate that all or part of step in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium may include：ROM, RAM, disk or CD etc..
Although present disclosure is as above, present invention is not limited to this.Any those skilled in the art are not departing from this
It in the spirit and scope of invention, can make various changes or modifications, therefore protection scope of the present invention should be with claim institute
Subject to the range of restriction.
Claims (20)
1. a kind of prediction technique of order, which is characterized in that including：
Obtain the corresponding time series of History Order；
Based on the time series, characteristic sequence is extracted；
Based on the characteristic sequence, 2D signal characteristic pattern is generated；
Based on the 2D signal characteristic pattern, neural network model is built, and ordered according to constructed neural network model
Single prediction.
2. the prediction technique of order according to claim 1, which is characterized in that the corresponding time series of the acquisition order
Including：
Obtain the corresponding initial data of History Order；
Initial data is pre-processed, the corresponding time series of order is obtained.
3. the prediction technique of order according to claim 2, which is characterized in that the pretreatment includes following at least one
Kind：Outlier processing, missing values processing.
4. the prediction technique of order according to claim 1, which is characterized in that the extraction characteristic sequence includes：It is based on
Wavelet Transformation Algorithm extracts characteristic sequence.
5. the prediction technique of order according to claim 1, which is characterized in that the generation 2D signal characteristic pattern packet
It includes：
Each characteristic sequence is divided into the sequence fragment that multiple length are n, wherein n is positive integer；
The corresponding sequence fragment of different characteristic sequence is replicated line by line, generates m row sequence fragments, the m rows sequence fragment
It is adjacent two-by-two in the ranks to meet different characteristic sequences, wherein m is positive integer；
Based on the m rows sequence fragment, the 2D signal characteristic pattern of m*n is generated.
6. the prediction technique of order according to claim 5, which is characterized in that it is described each characteristic sequence is divided into it is more
A length is that the sequence fragment of n includes：
Based on shifting function, each characteristic sequence is divided into the sequence fragment that multiple length are n.
7. the prediction technique of order according to claim 1, which is characterized in that the structure neural network model includes：
Build neural network model；
Based on the 2D signal characteristic pattern, the training neural network model obtains the neural network model parameter.
8. the prediction technique of order according to claim 7, which is characterized in that after building neural network model, also
Including：
Obtain the online data of order；
Based on online data training and update the neural network model.
9. the prediction technique of order according to claim 7 or 8, which is characterized in that the neural network is：Convolutional Neural
Network.
10. a kind of prediction meanss of order, which is characterized in that including：
First acquisition unit is suitable for obtaining the corresponding time series of History Order；
Extraction unit is suitable for being based on the time series, extracts characteristic sequence；
Generation unit is suitable for being based on the characteristic sequence, generates 2D signal characteristic pattern；
Construction unit is suitable for being based on the 2D signal characteristic pattern, builds neural network model, and according to constructed nerve net
Network model carries out order forecasting.
11. the prediction meanss of order according to claim 10, which is characterized in that the first acquisition unit includes：
First obtains subelement, is suitable for obtaining the corresponding initial data of History Order；
Second obtains subelement, suitable for being pre-processed to initial data, obtains the corresponding time series of order.
12. the prediction meanss of order according to claim 11, which is characterized in that the pretreatment includes following at least one
Kind：Outlier processing, missing values processing.
13. the prediction meanss of order according to claim 10, which is characterized in that the extraction unit is suitable for based on small
Wave conversion algorithm extracts characteristic sequence.
14. the prediction meanss of order according to claim 10, which is characterized in that the generation unit includes：
Divide subelement, suitable for each characteristic sequence is divided into the sequence fragment that multiple length are n, wherein n is positive integer；
Subelement is replicated, suitable for the corresponding sequence fragment of different characteristic sequence is replicated line by line, generation m row sequence fragments,
It is adjacent two-by-two in the ranks that the m rows sequence fragment meets different characteristic sequences, and wherein m is positive integer；
Subelement is generated, is suitable for being based on the m rows sequence fragment, generates the 2D signal characteristic pattern of m*n.
15. the prediction meanss of order according to claim 14, which is characterized in that the segmentation subelement, suitable for being based on
Each characteristic sequence is divided into the sequence fragment that multiple length are n by shifting function.
16. the prediction meanss of order according to claim 10, which is characterized in that the construction unit includes：
Subelement is built, is suitable for being based on the neural network model parameter, builds neural network model；
Training subelement, is suitable for being based on the 2D signal characteristic pattern, and the training neural network model obtains the nerve net
Network model parameter；
Subelement is predicted, suitable for carrying out order forecasting according to the neural network model trained.
17. the prediction meanss of order according to claim 16, which is characterized in that further include：
Second acquisition unit is suitable for obtaining the online data of order；
Updating unit, suitable for being based on the online data training and updating the neural network model.
18. the prediction meanss of order according to claim 16 or 17, which is characterized in that the neural network is：Convolution
Neural network.
19. a kind of computer readable storage medium, is stored thereon with computer instruction, which is characterized in that the computer instruction
Perform claim requires the step of any one of 1 to 9 the method when operation.
20. a kind of logistics system, including memory and processor, it is stored with and can runs on the processor on the memory
Computer instruction, which is characterized in that perform claim is any in requiring 1 to 9 when the processor runs the computer instruction
The step of item the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810357890.XA CN108564326B (en) | 2018-04-19 | 2018-04-19 | Order prediction method and device, computer readable medium and logistics system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810357890.XA CN108564326B (en) | 2018-04-19 | 2018-04-19 | Order prediction method and device, computer readable medium and logistics system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108564326A true CN108564326A (en) | 2018-09-21 |
| CN108564326B CN108564326B (en) | 2021-12-21 |
Family
ID=63535754
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810357890.XA Active CN108564326B (en) | 2018-04-19 | 2018-04-19 | Order prediction method and device, computer readable medium and logistics system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108564326B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109543924A (en) * | 2018-12-20 | 2019-03-29 | 上海德启信息科技有限公司 | Goods amount prediction technique, device and computer equipment |
| CN109685276A (en) * | 2018-12-27 | 2019-04-26 | 拉扎斯网络科技（上海）有限公司 | Order processing method, apparatus, electronic equipment and computer readable storage medium |
| CN110084437A (en) * | 2019-05-09 | 2019-08-02 | 上汽安吉物流股份有限公司 | Prediction technique and device, the logistics system and computer-readable medium of order |
| CN110084438A (en) * | 2019-05-09 | 2019-08-02 | 上汽安吉物流股份有限公司 | Prediction technique and device, the logistics system and computer-readable medium of order |
| CN110097320A (en) * | 2019-05-09 | 2019-08-06 | 上汽安吉物流股份有限公司 | Order forecast method and device, logistics system and computer-readable medium |
| CN110110931A (en) * | 2019-05-09 | 2019-08-09 | 上汽安吉物流股份有限公司 | Order forecast method and device, logistics system and computer-readable medium |
| CN110110932A (en) * | 2019-05-09 | 2019-08-09 | 上汽安吉物流股份有限公司 | Order forecast method and device, logistics system and computer-readable medium |
| CN110309947A (en) * | 2019-05-09 | 2019-10-08 | 上汽安吉物流股份有限公司 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
| CN110309948A (en) * | 2019-05-09 | 2019-10-08 | 上汽安吉物流股份有限公司 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
| CN110458664A (en) * | 2019-08-06 | 2019-11-15 | 上海新共赢信息科技有限公司 | A kind of user's trip information prediction technique, device, equipment and storage medium |
| CN110689278A (en) * | 2019-10-11 | 2020-01-14 | 珠海格力电器股份有限公司 | Sheet metal material management method and system, storage medium and electronic equipment |
| CN111461815A (en) * | 2020-03-17 | 2020-07-28 | 上海携程国际旅行社有限公司 | Order recognition model generation method, recognition method, system, device and medium |
| CN113435968A (en) * | 2021-06-23 | 2021-09-24 | 南京领行科技股份有限公司 | Network appointment vehicle dispatching method, device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105373840A (en) * | 2015-10-14 | 2016-03-02 | 深圳市天行家科技有限公司 | Designated-driving order predicting method and designated-driving transport capacity scheduling method |
| US20160071010A1 (en) * | 2014-05-31 | 2016-03-10 | Huawei Technologies Co., Ltd. | Data Category Identification Method and Apparatus Based on Deep Neural Network |
| CN107239532A (en) * | 2017-05-31 | 2017-10-10 | 北京京东尚科信息技术有限公司 | Data digging method and device |
| CN107679462A (en) * | 2017-09-13 | 2018-02-09 | 哈尔滨工业大学深圳研究生院 | A kind of depth multiple features fusion sorting technique based on small echo |
| CN107909206A (en) * | 2017-11-15 | 2018-04-13 | 电子科技大学 | A kind of PM2.5 Forecasting Methodologies based on deep structure Recognition with Recurrent Neural Network |
-
2018
- 2018-04-19 CN CN201810357890.XA patent/CN108564326B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160071010A1 (en) * | 2014-05-31 | 2016-03-10 | Huawei Technologies Co., Ltd. | Data Category Identification Method and Apparatus Based on Deep Neural Network |
| CN105373840A (en) * | 2015-10-14 | 2016-03-02 | 深圳市天行家科技有限公司 | Designated-driving order predicting method and designated-driving transport capacity scheduling method |
| CN107239532A (en) * | 2017-05-31 | 2017-10-10 | 北京京东尚科信息技术有限公司 | Data digging method and device |
| CN107679462A (en) * | 2017-09-13 | 2018-02-09 | 哈尔滨工业大学深圳研究生院 | A kind of depth multiple features fusion sorting technique based on small echo |
| CN107909206A (en) * | 2017-11-15 | 2018-04-13 | 电子科技大学 | A kind of PM2.5 Forecasting Methodologies based on deep structure Recognition with Recurrent Neural Network |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109543924A (en) * | 2018-12-20 | 2019-03-29 | 上海德启信息科技有限公司 | Goods amount prediction technique, device and computer equipment |
| CN109685276A (en) * | 2018-12-27 | 2019-04-26 | 拉扎斯网络科技（上海）有限公司 | Order processing method, apparatus, electronic equipment and computer readable storage medium |
| CN110110932A (en) * | 2019-05-09 | 2019-08-09 | 上汽安吉物流股份有限公司 | Order forecast method and device, logistics system and computer-readable medium |
| CN110084438A (en) * | 2019-05-09 | 2019-08-02 | 上汽安吉物流股份有限公司 | Prediction technique and device, the logistics system and computer-readable medium of order |
| CN110097320A (en) * | 2019-05-09 | 2019-08-06 | 上汽安吉物流股份有限公司 | Order forecast method and device, logistics system and computer-readable medium |
| CN110110931A (en) * | 2019-05-09 | 2019-08-09 | 上汽安吉物流股份有限公司 | Order forecast method and device, logistics system and computer-readable medium |
| CN110084437A (en) * | 2019-05-09 | 2019-08-02 | 上汽安吉物流股份有限公司 | Prediction technique and device, the logistics system and computer-readable medium of order |
| CN110309947A (en) * | 2019-05-09 | 2019-10-08 | 上汽安吉物流股份有限公司 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
| CN110309948A (en) * | 2019-05-09 | 2019-10-08 | 上汽安吉物流股份有限公司 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
| CN110458664A (en) * | 2019-08-06 | 2019-11-15 | 上海新共赢信息科技有限公司 | A kind of user's trip information prediction technique, device, equipment and storage medium |
| CN110689278A (en) * | 2019-10-11 | 2020-01-14 | 珠海格力电器股份有限公司 | Sheet metal material management method and system, storage medium and electronic equipment |
| CN111461815A (en) * | 2020-03-17 | 2020-07-28 | 上海携程国际旅行社有限公司 | Order recognition model generation method, recognition method, system, device and medium |
| CN111461815B (en) * | 2020-03-17 | 2023-04-28 | 上海携程国际旅行社有限公司 | Order recognition model generation method, recognition method, system, equipment and medium |
| CN113435968A (en) * | 2021-06-23 | 2021-09-24 | 南京领行科技股份有限公司 | Network appointment vehicle dispatching method, device, electronic equipment and storage medium |
| CN113435968B (en) * | 2021-06-23 | 2024-02-02 | 南京领行科技股份有限公司 | Network appointment vehicle dispatching method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108564326B (en) | 2021-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108564326A (en) | Prediction technique and device, computer-readable medium, the logistics system of order | |
| CN111241952B (en) | Reinforced learning reward self-learning method in discrete manufacturing scene | |
| Al-Dahidi et al. | Remaining useful life estimation in heterogeneous fleets working under variable operating conditions | |
| CN107133181B (en) | A kind of construction method of difference wavelet neural network software fault prediction technology | |
| CN103676649A (en) | Local self-adaptive WNN (Wavelet Neural Network) training system, device and method | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| EP3502978A1 (en) | Meta-learning system | |
| CN114463540A (en) | Segmenting images using neural networks | |
| CN112529053A (en) | Short-term prediction method and system for time sequence data in server | |
| CN109558898B (en) | Multi-choice learning method with high confidence based on deep neural network | |
| CN110740063B (en) | Network flow characteristic index prediction method based on signal decomposition and periodic characteristics | |
| Akbari et al. | Similarity-based error prediction approach for real-time inflow forecasting | |
| CN116822742A (en) | Power load prediction method based on dynamic decomposition-reconstruction integrated processing | |
| CN109800866B (en) | Reliability increase prediction method based on GA-Elman neural network | |
| CN115081609A (en) | Acceleration method in intelligent decision, terminal equipment and storage medium | |
| CN115619028A (en) | Clustering algorithm fusion-based power load accurate prediction method | |
| CN112667394B (en) | Computer resource utilization rate optimization method | |
| CN115222773A (en) | Single-point motion learning method and device | |
| CN114897274A (en) | Method and system for improving time sequence prediction effect | |
| CN115907000A (en) | Small sample learning method for optimal power flow prediction of power system | |
| Niina et al. | The spherical hidden Markov self organizing map for learning time series data | |
| CN114720129A (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
| CN114463994A (en) | Chaos and reinforcement learning based traffic flow prediction parallel method | |
| CN114492988A (en) | Method and device for predicting product yield in catalytic cracking process | |
| Li et al. | A novel interacting multiple-model method and its application to moisture content prediction of ASP flooding |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |