CN106952060A - Logistic Scheduling method and device - Google Patents
Logistic Scheduling method and device Download PDFInfo
- Publication number
- CN106952060A CN106952060A CN201610633455.6A CN201610633455A CN106952060A CN 106952060 A CN106952060 A CN 106952060A CN 201610633455 A CN201610633455 A CN 201610633455A CN 106952060 A CN106952060 A CN 106952060A
- Authority
- CN
- China
- Prior art keywords
- machine learning
- learning model
- logistic
- model
- 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 42
- 238000010801 machine learning Methods 0.000 claims abstract description 156
- 238000012549 training Methods 0.000 claims abstract description 33
- 235000013399 edible fruits Nutrition 0.000 claims description 2
- 230000008569 process Effects 0.000 description 9
- 238000004088 simulation Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000006467 substitution reaction Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000010009 beating Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
-
- 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
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the present invention is that, on a kind of Logistic Scheduling method and device, its method includes：Obtain the first regular scoring model in the logistic dispatching system；The regular scoring model of default machine learning model training fitting described first is called, the first machine learning model is obtained；First machine learning model is trained by default History Order data, the second machine learning model is obtained；It regard second machine learning model as the regular scoring model in the logistic dispatching system.So by continuous iteration, until obtained the second machine learning model tended towards stability, and call the second machine learning model to each related marking link marking in logistic dispatching system, marking accuracy can be greatly improved, and then improve logistics distribution efficiency.
Description
Technical field
The present embodiments relate to logistlcs technology field, more particularly to a kind of Logistic Scheduling method and device.
Background technology
With continuing to develop for society, pass through O2O (on Online To Offline, line to line under) platform transacting business
User it is also more and more.Many businessmans are all proposed app (Application, the application program) platform of oneself, and user passes through
Registered on related app, the registered user as the app, the registered user can just place an order on the app, such as order take-away,
Buy clothes or food etc..Businessman can assign the related person of sending with charge free to incite somebody to action after the order of user is connected to by logistic dispatching system
The article that places an order of user gives user.
In logistic dispatching system, there are many links, such as：Judge whether order adds be assigned to the link of the person of sending with charge free, order
Link etc. is associated between single packet link and order group and the person of sending with charge free.Logistic dispatching system needs to wait links to above-mentioned
Marking, and then order packet, the specified person of sending with charge free etc. are determined according to marking situation.However, due to currently existing logistic dispatching system
The accuracy given a mark to each link is not high, when the marking result progress logistics according to logistic dispatching system is sent with charge free so that logistics
Send with charge free less efficient.
The content of the invention
In order to improve the accuracy that existing logistic dispatching system is given a mark to each link so that according to logistic dispatching system
When marking result progress logistics is sent with charge free, improve logistics and send efficiency with charge free, the embodiment of the present invention provides a kind of Logistic Scheduling method and dress
Put.
First aspect according to embodiments of the present invention there is provided a kind of Logistic Scheduling method, including：
Obtain the first regular scoring model in the logistic dispatching system；
The regular scoring model of default machine learning model training fitting described first is called, the first machine learning mould is obtained
Type；
First machine learning model is trained by default History Order data, the second machine learning model is obtained；
It regard second machine learning model as the regular scoring model in the logistic dispatching system；
Determine that the logistic dispatching system dispenses the Logistic Scheduling of object to target according to the goal rule scoring model.
Second aspect according to embodiments of the present invention there is provided a kind of Logistic Scheduling device, including：
First regular scoring model acquiring unit, for obtaining the first rule marking mould in the logistic dispatching system
Type；
First machine learning model generation unit, for calling default machine learning model training fitting first rule
Scoring model, obtains the first machine learning model；
Second machine learning model generation unit, for by presetting History Order data to the first machine learning mould
Type training, obtains the second machine learning model；
First regular scoring model determining unit, for regarding second machine learning model as the Logistic Scheduling system
Regular scoring model in system；
Logistic Scheduling determining unit, for determining the logistic dispatching system to mesh according to the goal rule scoring model
Standard configuration send the Logistic Scheduling of object.
The technical scheme that embodiments of the invention are provided can include the following benefits：
Logistic Scheduling method and device provided in an embodiment of the present invention, the first rule in logistic dispatching system is got
After scoring model, call default machine learning model to be fitted the first regular scoring model, obtain the first machine learning model,
Training by the default History Order data that get to the first machine learning model, can obtain the second machine learning mould
Type, the second machine learning model is replaced the in logistic dispatching system first regular scoring model as in logistic dispatching system
Regular scoring model.So each related marking link in logistic dispatching system is beaten by the second obtained machine learning model
Point, marking accuracy can be greatly improved, and then improve logistics distribution efficiency.
It should be appreciated that the general description of the above and detailed description hereinafter are only exemplary and explanatory, not
The embodiment of the present invention can be limited.
Brief description of the drawings
Accompanying drawing herein is merged in specification and constitutes the part of this specification, shows the implementation for meeting the present invention
Example, and be used to together with specification to explain the principle of the embodiment of the present invention.
Fig. 1 is a kind of flow chart of Logistic Scheduling method according to an exemplary embodiment of the invention；
Fig. 2 is the flow chart of step S130 in Fig. 1；
Fig. 3 is the flow chart of step S136 in Fig. 2；
Fig. 4 is the flow chart of step S140 in Fig. 1；
Fig. 5 is a kind of flow chart of Logistic Scheduling method according to an exemplary embodiment of the invention；
Fig. 6 is the flow chart of step S150 in Fig. 1；
Fig. 7 is a kind of structural representation of Logistic Scheduling device according to an exemplary embodiment of the invention；
Fig. 8 is the schematic diagram of the second machine learning model generation unit in Fig. 7；
Fig. 9 is the schematic diagram of the first machine learning model training module in Fig. 8；
Figure 10 is the schematic diagram of goal rule scoring model determining unit in Fig. 7；
Figure 11 is a kind of structural representation of Logistic Scheduling device according to an exemplary embodiment of the invention；
Figure 12 is the schematic diagram of Logistic Scheduling determining unit in Fig. 7.
Embodiment
Here exemplary embodiment will be illustrated in detail, its example is illustrated in the accompanying drawings.Following description is related to
During accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represent same or analogous key element.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the embodiment of the present invention.On the contrary, they be only with
As be described in detail in the appended claims, embodiment of the present invention some in terms of consistent apparatus and method example.
Lower logistic dispatching system is briefly introduced first below, it is exemplary, after user places an order on O2O platforms, logistics
Scheduling system can obtain the sequence information of user, and according to the sequence information, judge whether to need to add the order to
Send the knight (i.e. the person of sending with charge free) of goods with charge free；If it did not, being so grouped to the order.For example, same type of order is divided
For one group, it is easy to knight's uniform dispatch.
Logistic dispatching system can run into all many links described above, in order to ensure in user's order dealing process
The timely of order is sent with charge free, it is necessary to the links marking in logistic dispatching system, typically call the marking of logistic dispatching system
Rule is given a mark to each link in logistic dispatching system, and according to marking result, the high result of selection marking arranges logistics to send with charge free.
Exemplary, following links in logistic dispatching system can be given a mark, such as：To whether needing to enter order addition to knight
Row marking；According to marking result, marking is chosen more than certain numerical value and marking highest knight, if it did not, so just need not
By the order addition to knight.Order packet can also be given a mark；According to marking result, the order is divided into marking knot
Fruit highest order group, is easy to the uniform dispatch of knight.Order group and knight can also be given a mark；According to marking result, choose
Order group correspondence marking highest knight, illustrates that the knight is best suitable for sending with charge free for the order group.Certainly, except marking can be used
As a result in, choose outside marking result highest, in other modes, for example, can respectively be given a mark to order group, knight, select
The progress close with knight's fraction of order group is matched etc..
After above-mentioned logistic dispatching system and marking rule is simply introduced, in order to solve currently existing logistic dispatching system pair
The accuracy of each link marking is not high, causes when the marking result progress logistics according to logistic dispatching system is sent with charge free so that thing
The problem of school send less efficient, the embodiment of the present invention provide firstly a kind of Logistic Scheduling method, as shown in figure 1, this method
It may include steps of：
In step s 110, the first regular scoring model in logistic dispatching system is obtained.
In logistic dispatching system, due to needing to give a mark to many links therein, it is easy to be ordered according to marking result
Single logistics is sent with charge free, therefore is had in logistic dispatching system in the marking rule model of itself, embodiment and be referred to as the scoring model
First regular scoring model.The first regular scoring model, is to call relevant criterion, related link is given a mark, for example：To ordering
In the marking link of single packet, according to sequence information, the cargo type that is dispensed in such as the order, dispatching place, dispatching requirement,
The order packet that selection and cargo type, dispatching place, dispatching requirement etc. match, then the marking result that the order is grouped exists
In all related order packets, marking should be highest.
But, many times, the original marking rule model (the i.e. first regular scoring model) in logistic dispatching system is right
The marking of correlation marking link is not accurately fixed high, and the order particularly in user is complicated, user's blanket order amount is violent in certain period
Increase etc., the goods of multiple species is included in such as order, it is desirable to which means of distribution has when particular/special requirement, and violent in order volume
In the case of increasing, situations such as marking is disorderly occurs in the original marking rule model in logistic dispatching system, in turn results in logistics tune
Degree system can not be sent with charge free according to goods in the marking result reasonable arrangement order of the first regular scoring model so that logistics distribution
Efficiency is substantially reduced.
In the step s 120, the regular scoring model of default machine learning model training fitting first is called, the first machine is obtained
Device learning model.
Default machine learning model in the embodiment of the present invention, can be existing DNN (Deep Neural
Networks, deep neural network), Logic Regression Models etc. are can also be, can be chosen as needed, the embodiment of the present invention
Not limited to this.
Due to having in the first regular scoring model to scoring criterion for link of being given a mark in logistic dispatching system etc., it is therefore desirable to
The regular scoring model of default machine learning model training fitting first is called, the first machine learning model is so obtained.This first
Machine learning model, not only with the scoring criterion in the first regular scoring model, and also has default machine learning model
Mode classification, can improve the marking accuracy for link of being given a mark in logistic dispatching system.
In step s 130, the first machine learning model is trained by default History Order data, obtains the second machine
Learning model.
After user places an order, because logistic dispatching system can be given a mark by the first regular scoring model to each marking link,
Therefore the History Order data in logistic dispatching system can be collected, in History Order data, the feature comprising order group, ridden
Feature of scholar etc., and order group characteristic value and the characteristic value etc. of knight.If the History Order number in logistic dispatching system
According to more, the partial history order data needed for it can be filtered out as default History Order data.And it is default by this
History Order data are trained to the first machine learning model, obtain the second machine learning model.
In step S140, the second machine learning model is regard as the regular scoring model in logistic dispatching system.
In step S150, determine that logistic dispatching system dispenses the logistics of object to target according to goal rule scoring model
Scheduling.
Second machine learning model obtained above, not only beating with the in logistic dispatching system first regular scoring model
Minute mark standard etc., but also with the identification mode classification in the first machine learning model, and by presetting History Order data
To the first machine learning model train, the second obtained machine learning model to logistic dispatching system marking link marking when,
It can greatly improve the marking accuracy of logistic dispatching system, and then improve logistics and send efficiency with charge free, therefore by the second machine learning
Model replaces the in logistic dispatching system first regular scoring model as the regular scoring model of logistic dispatching system, to logistics
Each marking link marking in scheduling system.
Logistic Scheduling method provided in an embodiment of the present invention, the first rule marking mould in logistic dispatching system is got
After type, call default machine learning model to be fitted the first regular scoring model, the first machine learning model is obtained, by obtaining
Training of the default History Order data got to the first machine learning model, can obtain the second machine learning model, by
The first regular scoring model in two machine learning models substitution logistic dispatching system is beaten as the rule in logistic dispatching system
Sub-model.So each related marking link in logistic dispatching system is given a mark by obtained the second machine learning model, can be with
Marking accuracy is greatly improved, and then improves logistics distribution efficiency.
In order to elaborate on how that the first machine learning model is trained by default History Order data, Fig. 1 side is used as
The refinement of method, in another embodiment of the invention, as shown in Fig. 2 step S130 can also comprise the following steps：
In step S131, the History Order data of logistic dispatching system are obtained.
The History Order data are that logistic dispatching system can be given birth to by the first regular scoring model to each marking link marking
Into, in History Order data, feature of feature, knight comprising order group etc., and order group characteristic value and knight
Characteristic value etc..
In step S132, first marking result of the first regular scoring model to History Order data is obtained.
In History Order data, the first rule marking mould in the order placement information of user, logistic dispatching system is often received
Type all can related marking link marking, obtain result of giving a mark., will be by the first regular scoring model to going through in the embodiment of the present invention
The related marking result of history order, the referred to as first marking result.
In step S133, History Order data are given a mark by the first machine learning model, the second marking result is obtained.
In the embodiment of the present invention, History Order data are given a mark by the first machine learning model, referred to the first machine
Learning model replaces the in logistic dispatching system first regular scoring model, each order numbers in History Order data
According to respectively being given a mark in logistic dispatching system, link is given a mark, and obtained marking result is used as to the second marking result.
It should be noted that in the other embodiment that the present invention is provided, can also be based on logistic dispatching system, obtaining
A set of simulation logistic dispatching system, by History Order in the simulation logistic dispatching system operation simulation, finally calculate all
The indices of order, such as：Averagely dispatching duration, most slow 10% order averagely dispense duration, dispense punctual rate and free travel distance
Deng according to circumstances setting each index weights, a catalogue scale value can be obtained.
In step S134, judge whether the first marking result is identical with the second marking result.
First marking result with second marking result it is variant when, in step S135, obtain first give a mark result with
Different difference marking result between second marking result.
In step S136, the target histories order numbers corresponding with difference marking result are obtained in History Order data
According to.
In step S137, the first machine learning model is trained by target histories order data.
If the marking of the first marking result and the second marking result is variant, then obtain discrepant difference in the two
Marking result.The corresponding target histories order data of difference marking result is obtained again, passes through the target histories order data pair
First machine learning model is trained.
As the refinement of Fig. 2 methods, in the another embodiment that the present invention is provided, as shown in figure 3, step S136 can be with
Comprise the following steps：
In step S1361, in target histories order, by the marking result of the first machine learning model higher than the first rule
Then the order of scoring model marking result beats the marking result of the first machine learning model less than the first rule as positive sample
The order of sub-model marking result is used as negative sample.
In step S1362, the first machine learning model is trained by positive sample and negative sample respectively.
So initial machine learning model is trained by obtained positive sample and negative sample respectively, the second machine can be obtained
Device learning model so that after the first regular scoring model during the second machine learning model to be replaced to logistic dispatching system, make
The marking accuracy of each marking link in logistic dispatching system must be greatly improved by the second machine learning model, Jin Erti
High logistics distribution efficiency.
Based on the refinement of Fig. 1 methods, in the another embodiment that the present invention is provided, as shown in figure 4, step S140 can be with
Comprise the following steps：
In step s 11, it regard the second machine learning model as the goal rule scoring model in logistic dispatching system.
Or,
In step s 141, the positive sample and negative sample in default History Order data are obtained respectively.
On the basis of above-described embodiment, due to the first machine learning model and the second machine learning model, respectively to going through
Order data in history order to each marking link beat in logistic dispatching system (or simulation logistic dispatching system)
Point, obtain result of giving a mark.In the way of above-described embodiment is taken, extract the two marking and the difference marking result of difference occur,
And the corresponding History Order data of difference marking result are obtained, the marking of the second machine learning model is higher than the first machine learning
The History Order data of model are used as positive sample；Second machine learning model is given a mark less than the history of the first machine learning model
Order data is used as negative sample.
In step S142, the second machine learning model is trained by positive sample and negative sample, the 3rd engineering is obtained
Practise model.
In step S143, the 3rd machine learning model is regard as the regular scoring model in logistic dispatching system.
By positive sample obtained above and negative sample, the second machine learning model is trained respectively, to reach that optimization should
The purpose of second machine learning, obtains the 3rd machine learning model, using the 3rd machine learning model as in logistic dispatching system
Regular scoring model can further improve the accuracy of marking, and then improve logistics distribution efficiency.According in this way, this hair
In bright embodiment, positive sample and negative sample can be constantly obtained, by positive sample and negative sample constantly to machine learning model
Training, continues to optimize machine learning model so that the accuracy more and more higher of obtained machine learning model marking.
As the refinement of Fig. 1 methods, in the another embodiment that the present invention is provided, as shown in figure 5, also including default machine
The acquisition modes of learning model, i.e. this method can also comprise the following steps：
In step S180, the History Order data of logistic dispatching system are obtained.
In step S190, default machine learning model is trained by History Order data.
Before step S120, in addition it is also necessary to obtain the History Order data of logistic dispatching system, ordered with will pass through the history
Forms data is to default machine learning model training, so that the default machine learning model can be realized to phase in logistic dispatching system
Close the marking of marking link.
In addition, as the refinement of Fig. 1 methods, in the present invention provides another embodiment, as shown in fig. 6, step S150 is also
It may include steps of：
In step S151, the dispatching parameter information that target dispenses object is obtained.
The dispatching parameter information, can be that the volumes of sending objects, weight, dispatching place, dispatching are required etc..
In step S152, according to dispatching parameter information and goal rule scoring model to each in logistic dispatching system
Link marking is dispensed, it is determined that dispatching link combination.
For example, can will be currently at idle condition to dispatching personnel's marking and be adapted to make a gift to someone matching somebody with somebody with the dispatching object
Member's marking is most high.
In step S153, object progress Logistic Scheduling is dispensed to target by dispensing link combination.
According to it is each dispatching link marking result, choose marking highest as target dispense object dispatching combine into
Row dispatching.
The part can specifically participate in above-described embodiment, repeat no more here.
Logistic Scheduling method provided in an embodiment of the present invention, the first rule marking mould in logistic dispatching system is got
After type, call default machine learning model to be fitted the first regular scoring model, the first machine learning model is obtained, by obtaining
Training of the default History Order data got to the first machine learning model, can obtain the second machine learning model, by
The first regular scoring model in two machine learning models substitution logistic dispatching system is beaten as the rule in logistic dispatching system
Sub-model.So each related marking link in logistic dispatching system is given a mark by obtained the second machine learning model, can be with
Marking accuracy is greatly improved, and then improves logistics distribution efficiency.
And because the first regular scoring model in existing logistic dispatching system is only in that the coefficient of training pattern, and
And each system system is required for running a simulation system, so as to determine to retain or give up, there is parameter and choose not enough and training consumption
When it is excessive the problems such as；And inventive embodiments can so be easy to be collected into greatly by section regression scoring model of giving a mark
Sample is measured, while training cost relatively low, and more complicated model can be trained；Each version mould is run by simulation system again
Type, then sample is collected, more preferable model is further trained, continuous iteration is until tending towards stability, by this way, trains
Model allows scheduling system more to level off to optimal solution.
The description of embodiment of the method more than, it is apparent to those skilled in the art that the present invention is real
Applying example can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but many situations
It is lower the former be more preferably embodiment.Understood based on such, the technical scheme of the embodiment of the present invention is substantially in other words to existing
The part for having technology to contribute can be embodied in the form of software product, and the computer software product is stored in one and deposited
In storage media, including some instructions are make it that a computer equipment (can be personal computer, server, or network
Equipment etc.) perform all or part of step of each of the invention embodiment methods described.And foregoing storage medium includes：It is read-only
Memory (ROM), random access memory (RAM), magnetic disc or CD etc. are various can be with the medium of store program codes.
In addition, as the realization to the various embodiments described above, the embodiment of the present invention additionally provides a kind of Logistic Scheduling device, should
Device is located in server, as shown in fig. 7, the device includes：
First regular scoring model acquiring unit 10, for obtaining the first rule marking mould in the logistic dispatching system
Type；
First machine learning model generation unit 20, for calling default machine learning model training fitting first rule
Then scoring model, obtains the first machine learning model；
Second machine learning model generation unit 30, for by presetting History Order data to first machine learning
Model training, obtains the second machine learning model；
Goal rule scoring model determining unit 40, for regarding second machine learning model as the Logistic Scheduling
Regular scoring model in system.
Logistic Scheduling determining unit 50, for determining the logistic dispatching system pair according to the goal rule scoring model
Target dispenses the Logistic Scheduling of object.
In still another embodiment of the process, based on Fig. 7, as shown in figure 8, the second machine learning model generation unit
30, including：
History Order data acquisition module 31, the History Order data for obtaining the logistic dispatching system；
First marking result acquisition module 32, for obtaining the described first regular scoring model to the History Order data
First marking result；
Second marking result acquisition module 33, for by first machine learning model to the History Order data
Marking, obtains the second marking result；
Give a mark result judge module 34, for judge the first marking result with described second give a mark result whether phase
Together；
Difference marking result acquisition module 35, for variant in the described first marking result and the described second marking result
When, obtain the difference marking result between the first marking result and the second marking result；
Target histories order data acquisition module 36, gives a mark for being obtained in the History Order data with the difference
As a result corresponding target histories order data；
First machine learning model training module 37, for by the target histories order data to first machine
Learning model is trained.
In still another embodiment of the process, based on Fig. 8, as shown in figure 9, the first machine learning model training module
36, including：
Positive sample determination sub-module 361, in the target histories order, by first machine learning model
Marking result is used as positive sample higher than the order of the described first regular scoring model marking result；
Negative sample determination sub-module 362, for the marking result of first machine learning model to be less than into described first
The order of regular scoring model marking result is used as negative sample；
First machine learning model trains submodule 363, for respectively by the positive sample and the negative sample to institute
State the training of the first machine learning model.
In still another embodiment of the process, based on Fig. 7, as shown in Figure 10, the goal rule scoring model determining unit
40 include：
First object rule scoring model determining module 411, for regarding second machine learning model as the thing
Flow the goal rule scoring model in scheduling system；Or,
Sample acquisition module 41, for obtaining positive sample and negative sample in the default History Order data respectively；
Second machine learning model acquisition module 42, for by the positive sample and the negative sample to second machine
Device learning model is trained, and obtains the second machine learning model；
Second Rule scoring model determining module 43, for regarding second machine learning model as the Logistic Scheduling
Regular scoring model in system.
In still another embodiment of the process, based on Fig. 7, as shown in figure 11, the device also includes：
History Order data cell 80, the History Order data for obtaining the logistic dispatching system；
Default machine learning model training unit 90, for by the History Order data to the default machine learning
Model training.
In still another embodiment of the process, based on Fig. 7, as shown in figure 12, Logistic Scheduling determining unit 50 can also be wrapped
Include：
Parameter information acquisition module 51, the dispatching parameter information of object is dispensed for obtaining target；
Link combination determining module 52 is dispensed, for according to the dispatching parameter information and goal rule marking mould
Type is given a mark to each dispatching link in the logistic dispatching system, it is determined that dispatching link combination；
Logistics distribution module 53, logistics tune is carried out for dispensing object to the target by the dispatching link combination
Degree.
On the device in above-described embodiment, wherein modules perform the concrete mode of operation in relevant this method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
Logistic Scheduling device provided in an embodiment of the present invention, the first rule marking mould in logistic dispatching system is got
After type, call default machine learning model to be fitted the first regular scoring model, the first machine learning model is obtained, by obtaining
Training of the default History Order data got to the first machine learning model, can obtain the second machine learning model, by
The first regular scoring model in two machine learning models substitution logistic dispatching system is beaten as the rule in logistic dispatching system
Sub-model.So each related marking link in logistic dispatching system is given a mark by obtained the second machine learning model, can be with
Marking accuracy is greatly improved, and then improves logistics distribution efficiency.
And because the first regular scoring model in existing logistic dispatching system is only in that the coefficient of training pattern, and
And each system system is required for running a simulation system, so as to determine to retain or give up, there is parameter and choose not enough and training consumption
When it is excessive the problems such as；And inventive embodiments can so be easy to be collected into greatly by section regression scoring model of giving a mark
Sample is measured, while training cost relatively low, and more complicated model can be trained；Each version mould is run by simulation system again
Type, then sample is collected, more preferable model is further trained, continuous iteration is until tending towards stability, by this way, trains
Model allows scheduling system more to level off to optimal solution.
It is understood that the embodiment of the present invention can be used in numerous general or special purpose computing system environments or configuration.
For example：Personal computer, server computer, handheld device or portable set, laptop device, multicomputer system, base
In the system of microprocessor, set top box, programmable consumer-elcetronics devices, network PC, minicom, mainframe computer, bag
Include DCE of any of the above system or equipment etc..
The embodiment of the present invention can be described in the general context of computer executable instructions, example
Such as program module.Usually, program module include performing particular task or realize the routine of particular abstract data type, program,
Object, component, data structure etc..The embodiment of the present invention can also be put into practice in a distributed computing environment, it is distributed at these
In computing environment, task is performed by the remote processing devices connected by communication network.In a distributed computing environment,
Program module can be located at including in the local and remote computer-readable storage medium including storage device.
It should be noted that herein, the relational terms of such as " first " and " second " or the like are used merely to one
Individual entity or operation make a distinction with another entity or operation, and not necessarily require or imply these entities or operate it
Between there is any this actual relation or order.Moreover, term " comprising ", "comprising" or its any other variant are intended to
Cover including for nonexcludability, so that process, method, article or equipment including a series of key elements not only include those
Key element, but also other key elements including being not expressly set out, or also include for this process, method, article or set
Standby intrinsic key element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that
Also there is other identical element in the process including the key element, method, article or equipment.
Those skilled in the art will readily occur to this hair after considering specification and putting into practice inventive embodiments disclosed herein
Other embodiments of bright embodiment.Any modification, purposes or the adaptability that the application is intended to the embodiment of the present invention become
Change, these modifications, purposes or adaptations follow the general principle of the embodiment of the present invention and including the embodiment of the present invention
Undocumented common knowledge or conventional techniques in the art.Description and embodiments be considered only as it is exemplary,
The true scope and spirit of the embodiment of the present invention are pointed out by following claim.
It should be appreciated that the accurate knot that the embodiment of the present invention is not limited to be described above and is shown in the drawings
Structure, and various modifications and changes can be being carried out without departing from the scope.The scope of the embodiment of the present invention is only by appended right
It is required that to limit.
Claims (12)
1. a kind of Logistic Scheduling method, applied to logistic dispatching system, it is characterised in that including：
Obtain the first regular scoring model in the logistic dispatching system；
The regular scoring model of default machine learning model training fitting described first is called, the first machine learning model is obtained；
First machine learning model is trained by default History Order data, the second machine learning model is obtained；
The goal rule scoring model in the logistic dispatching system is determined according to second machine learning model；
Determine that the logistic dispatching system dispenses the Logistic Scheduling of object to target according to the goal rule scoring model.
2. Logistic Scheduling method according to claim 1, it is characterised in that described by presetting History Order data to institute
The training of the first machine learning model is stated, including：
Obtain the History Order data of the logistic dispatching system；
Obtain first marking result of the described first regular scoring model to the History Order data；
The History Order data are given a mark by first machine learning model, the second marking result is obtained；
Judge whether the first marking result is identical with the described second marking result；
When the described first marking result is differed with the described second marking result, the first marking result and described the is obtained
Different difference marking result between two marking results；
The target histories order data corresponding with difference marking result is obtained in the History Order data；
First machine learning model is trained by the target histories order data.
3. Logistic Scheduling method according to claim 2, it is characterised in that described to pass through the target histories order data
First machine learning model is trained, including：
In the target histories order, the marking result of first machine learning model is given a mark higher than the described first rule
The order of model marking result is as positive sample, by the marking result of first machine learning model less than the described first rule
The order of scoring model marking result is used as negative sample；
First machine learning model is trained by the positive sample and the negative sample respectively.
4. Logistic Scheduling method according to claim 1, it is characterised in that described according to second machine learning model
The goal rule scoring model in the logistic dispatching system is determined, including：
It regard second machine learning model as the goal rule scoring model in the logistic dispatching system；
Or；The positive sample and negative sample in the default History Order data are obtained respectively；By the positive sample and described
Negative sample is trained to second machine learning model, obtains the 3rd machine learning model；By the 3rd machine learning model
It is used as the goal rule scoring model in the logistic dispatching system.
5. according to any described Logistic Scheduling method in Claims 1-4, it is characterised in that the default machine learning mould
The acquisition modes of type, including：
Obtain the History Order data of the logistic dispatching system；
The first machine learning model is trained by the History Order data, the default machine learning model is obtained.
6. according to any described Logistic Scheduling method in Claims 1-4, it is characterised in that described to be advised according to the target
Then scoring model determines that the logistic dispatching system dispenses the Logistic Scheduling of object to target, including：
Obtain the dispatching parameter information that target dispenses object；
According to the dispatching parameter information and the goal rule scoring model to each dispatching in the logistic dispatching system
Link is given a mark, it is determined that dispatching link combination；
Object is dispensed to the target by the dispatching link combination and carries out Logistic Scheduling.
7. a kind of Logistic Scheduling device, it is characterised in that including：
First regular scoring model acquiring unit, for obtaining the first regular scoring model in the logistic dispatching system；
First machine learning model generation unit, for calling default machine learning model training fitting the first rule marking
Model, obtains the first machine learning model；
Second machine learning model generation unit, for being instructed by default History Order data to first machine learning model
Practice, obtain the second machine learning model；
Goal rule scoring model determining unit, for determining the logistic dispatching system according to second machine learning model
In goal rule scoring model；
Logistic Scheduling determining unit, for determining that the logistic dispatching system is matched somebody with somebody to target according to the goal rule scoring model
Send the Logistic Scheduling of object.
8. Logistic Scheduling device according to claim 7, it is characterised in that the second machine learning model generation is single
Member, including：
History Order data acquisition module, the History Order data for obtaining the logistic dispatching system；
First marking result acquisition module, for obtaining the described first regular scoring model to the first of the History Order data
Marking result；
Second marking result acquisition module, for being given a mark by first machine learning model to the History Order data,
Obtain the second marking result；
Marking result judge module, for judging whether the first marking result is identical with the described second marking result；
Difference marking result acquisition module, for when the described first marking result is different from the described second marking result, obtaining
Different difference marking result between the first marking result and the described second marking result；
Target histories order data acquisition module, for being obtained in the History Order data and difference marking result phase
Corresponding target histories order data；
First machine learning model training module, for by the target histories order data to the first machine learning mould
Type training.
9. Logistic Scheduling device according to claim 8, it is characterised in that first machine learning model trains mould
Block, including：
Positive sample determination sub-module, in the target histories order, by the marking knot of first machine learning model
Fruit is used as positive sample higher than the order of the described first regular scoring model marking result；
Negative sample determination sub-module, for the marking result of first machine learning model to be given a mark less than the described first rule
The order of model marking result is used as negative sample；
First machine learning model trains submodule, for respectively by the positive sample and the negative sample to first machine
Device learning model is trained.
10. Logistic Scheduling device according to claim 7, it is characterised in that the goal rule scoring model determines single
Member, including：
First object rule scoring model determining module, for regarding second machine learning model as the Logistic Scheduling system
Goal rule scoring model in system；Or,
Sample acquisition module, for obtaining positive sample and negative sample in the default History Order data respectively；
Second machine learning model acquisition module, for by the positive sample and the negative sample to second machine learning
Model training, obtains the second machine learning model；
Second Rule scoring model determining module, for using second machine learning model as in the logistic dispatching system
Regular scoring model.
11. according to any described Logistic Scheduling device in claim 7 to 10, it is characterised in that also include：
History Order data cell, the History Order data for obtaining the logistic dispatching system；
Default machine learning model training unit, for being instructed by the History Order data to the default machine learning model
Practice.
12. according to any described Logistic Scheduling device in claim 7 to 10, it is characterised in that Logistic Scheduling determining unit,
Including：
Parameter information acquisition module, the dispatching parameter information of object is dispensed for obtaining target；
Dispense link combination determining module, for according to it is described dispatching parameter information and the goal rule scoring model to institute
Each dispatching link marking in logistic dispatching system is stated, it is determined that dispatching link combination；
Logistics distribution module, Logistic Scheduling is carried out for dispensing object to the target by the dispatching link combination.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610633455.6A CN106952060A (en) | 2016-08-04 | 2016-08-04 | Logistic Scheduling method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610633455.6A CN106952060A (en) | 2016-08-04 | 2016-08-04 | Logistic Scheduling method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106952060A true CN106952060A (en) | 2017-07-14 |
Family
ID=59466187
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610633455.6A Pending CN106952060A (en) | 2016-08-04 | 2016-08-04 | Logistic Scheduling method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106952060A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197825A (en) * | 2018-01-29 | 2018-06-22 | 北京小度信息科技有限公司 | System scheduling method and device |
| CN108335071A (en) * | 2018-02-06 | 2018-07-27 | 北京小度信息科技有限公司 | Dispense method for allocating tasks, device, electronic equipment and computer storage media |
| CN109447355A (en) * | 2018-10-31 | 2019-03-08 | 网易无尾熊（杭州）科技有限公司 | Dispatching optimization method, device, medium and the calculating equipment of articles from the storeroom |
| CN110874702A (en) * | 2018-09-04 | 2020-03-10 | 菜鸟智能物流控股有限公司 | Model training method and device in logistics sorting scene and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389639A (en) * | 2015-12-15 | 2016-03-09 | 上海汽车集团股份有限公司 | Logistics transportation route planning method, device and system based on machine learning |
| CN105468742A (en) * | 2015-11-25 | 2016-04-06 | 小米科技有限责任公司 | Malicious order recognition method and device |
| CN105632488A (en) * | 2016-02-23 | 2016-06-01 | 深圳市海云天教育测评有限公司 | Voice evaluation method and device |
| CN105741063A (en) * | 2016-01-21 | 2016-07-06 | 华南师范大学 | Personalized logistics scheduling method and personalized logistics scheduling system based on user's focus of attention |
-
2016
- 2016-08-04 CN CN201610633455.6A patent/CN106952060A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105468742A (en) * | 2015-11-25 | 2016-04-06 | 小米科技有限责任公司 | Malicious order recognition method and device |
| CN105389639A (en) * | 2015-12-15 | 2016-03-09 | 上海汽车集团股份有限公司 | Logistics transportation route planning method, device and system based on machine learning |
| CN105741063A (en) * | 2016-01-21 | 2016-07-06 | 华南师范大学 | Personalized logistics scheduling method and personalized logistics scheduling system based on user's focus of attention |
| CN105632488A (en) * | 2016-02-23 | 2016-06-01 | 深圳市海云天教育测评有限公司 | Voice evaluation method and device |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197825A (en) * | 2018-01-29 | 2018-06-22 | 北京小度信息科技有限公司 | System scheduling method and device |
| CN108335071A (en) * | 2018-02-06 | 2018-07-27 | 北京小度信息科技有限公司 | Dispense method for allocating tasks, device, electronic equipment and computer storage media |
| WO2019153761A1 (en) * | 2018-02-06 | 2019-08-15 | 北京小度信息科技有限公司 | Distribution task assignment method, apparatus, electronic device and computer storage medium |
| CN108335071B (en) * | 2018-02-06 | 2020-09-11 | 北京小度信息科技有限公司 | Distribution task allocation method and device, electronic equipment and computer storage medium |
| CN110874702A (en) * | 2018-09-04 | 2020-03-10 | 菜鸟智能物流控股有限公司 | Model training method and device in logistics sorting scene and electronic equipment |
| CN110874702B (en) * | 2018-09-04 | 2023-05-23 | 菜鸟智能物流控股有限公司 | Model training method and device under logistics sorting scene and electronic equipment |
| CN109447355A (en) * | 2018-10-31 | 2019-03-08 | 网易无尾熊（杭州）科技有限公司 | Dispatching optimization method, device, medium and the calculating equipment of articles from the storeroom |
| CN109447355B (en) * | 2018-10-31 | 2022-02-01 | 阿里巴巴（中国）有限公司 | Distribution optimization method, device, medium and computing equipment for warehouse goods |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liao et al. | Integrated FAHP, ARAS-F and MSGP methods for green supplier evaluation and selection | |
| CN107093084A (en) | Potential user predicts method for transformation and device | |
| CN106952060A (en) | Logistic Scheduling method and device | |
| CN107563705A (en) | Household electrical appliances product safety stock and the system and method ordered goods again are analyzed using big data | |
| CN107563841A (en) | A kind of commending system decomposed that scored based on user | |
| Eraslan | A decision making method via TOPSIS on soft sets | |
| CN110163647A (en) | A kind of data processing method and device | |
| CN107153977A (en) | Transaction entity credit estimation method, apparatus and system in online trade platform | |
| Paul et al. | An artificial neural network model for optimization of finished goods inventory | |
| CN109598430A (en) | Dispense range generation method, device, electronic equipment and storage medium | |
| Singh et al. | A closed loop supply chain system with flexible manufacturing and reverse logistics operation under shortages for deteriorating items | |
| CN107506845A (en) | A kind of electricity sales amount Forecasting Methodology and its system based on multi-model fusion | |
| CN109063935A (en) | A kind of method, apparatus and storage medium of prediction task processing time | |
| CN105574680A (en) | Diagnosis method and system of enterprise competitiveness | |
| CN107247666A (en) | A kind of feature based selection and the software defect number Forecasting Methodology of integrated study | |
| CN109118316A (en) | The recognition methods of shop authenticity and device on line | |
| Elanchezhian et al. | Vendor evaluation using multi criteria decision making | |
| Campana et al. | A proposal for the evaluation of craftsmanship in industry | |
| Li et al. | A knowledge discovery and reuse method for time estimation in ship block manufacturing planning using DEA | |
| CN110942180B (en) | Industrial design matching service side prediction method based on xgboost algorithm | |
| Torabi et al. | A possibilistic approach for designing hybrid cellular manufacturing systems | |
| EP2851851A1 (en) | A computer implemented tool and method for automating the forecasting process | |
| CN109784593A (en) | Production capacity equalization processing method and device for multistoried storehouse | |
| CN105930936B (en) | A kind of growth prediction method producing microorganism contained by food in the process of circulation | |
| Miller et al. | Interval type-2 fuzzy modelling and stochastic search for real-world inventory management |
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: 20170714 |
|
| RJ01 | Rejection of invention patent application after publication |