CN110309948A - Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium - Google Patents
Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium Download PDFInfo
- Publication number
- CN110309948A CN110309948A CN201910382944.2A CN201910382944A CN110309948A CN 110309948 A CN110309948 A CN 110309948A CN 201910382944 A CN201910382944 A CN 201910382944A CN 110309948 A CN110309948 A CN 110309948A
- Authority
- CN
- China
- Prior art keywords
- monthly
- complete vehicle
- vehicle logistics
- order
- model
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to logistics fields, more specifically, being related to a kind of monthly order forecast method of complete vehicle logistics and device, logistics system and computer-readable medium.This method comprises: obtaining the monthly complete vehicle logistics order data of history;Extract feature construction learning sample set;According to learning sample set, constructs Random Forest model and be trained and assess, obtain the monthly order forecasting model of complete vehicle logistics;The route predicted needed for input is predicted the following monthly complete vehicle logistics order of the route and is exported according to the monthly order forecasting model of complete vehicle logistics.The monthly order forecast method of complete vehicle logistics provided by the invention and device, system, storage medium, realize the prediction to the monthly order of complete vehicle logistics, greatly improve forecasting efficiency and prediction accuracy, for the pool transport power plan arrangement of carrier, storehouse management provides more accurately information, and then reduces production management cost.
Description
Technical field
The present invention relates to logistics fields, more specifically, be related to a kind of monthly order forecast method of complete vehicle logistics and device,
Logistics system and computer-readable medium.
Background technique
Complete vehicle logistics order it is monthly prediction be exactly by excavate historical information, construct prediction technique, to future it is monthly whole
Vehicle logistics order carries out Accurate Prediction, is overall arrangement carrier transport power, and warehousing management provides data reference.
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.
The monthly order forecasting of existing complete vehicle logistics is using manually according to History Order report, by history monthly order in the recent period
It is predicted with the history same period monthly average weighted mode, does so that one side efficiency is relatively low, and automated procedures are not high, separately
On the one hand, without the pattern information of sufficiently excavation History Order, prediction accuracy is not high.
Recently as the development of artificial intelligence technology, machine learning carries out in object identification, the fields such as price expectation
Successfully application.Currently, two most popular class machine learning algorithms are neural network algorithm and tree algorithm (random forest, ladder
Degree promotes decision tree etc.), the basis of tree algorithm is exactly decision tree.Decision tree is because of its readily understood, easy building, fireballing spy
Property, it is widely used in statistics, data mining, machine learning field.
Random forest (Random Forest, RF) is the representative algorithm of data mining thought, can be from limited data
Excavate a large amount of effective information.Random forests algorithm obtains training sample, base with bootstrap (Bootsrap) double sampling method
This thought is construction Multiple trees model, has precision of prediction is high, extensive error is controllable, fast convergence rate, adjustment parameter are few etc.
Advantage can effectively avoid the generation of " over-fitting " phenomenon, be particularly suitable for the operation of high dimensional data.
It is relevant there is presently no the related application for combining machine learning algorithm with complete vehicle logistics order forecasting
Order forecasting patent is distributed in e-commerce field more, if application publication number is right in the Chinese invention patent of CN104766144A
Daily order uses some rule-based predicting strategies, will classify on the date, and distinguishes the different classes of date
It is predicted, and is predicted in such a way that base period is multiplied by benchmark growth rate, there is no sufficiently excavate order volume wave among these
Dynamic pattern information.The Chinese invention patent that application publication number is CN104599002A is established using excavation order feature
The mapping relations of feature and order value predict order value.
Summary of the invention
The object of the present invention is to provide a kind of monthly order forecast method of complete vehicle logistics and device, system, storage medium, solutions
Intelligence degree is low, labor intensive is at high cost, forecasting efficiency is low in the certainly current monthly order forecasting of complete vehicle logistics and prediction is accurate
Spend the problem of difference.
To achieve the goals above, the present invention provides a kind of monthly order forecast methods of complete vehicle logistics, including following step
It is rapid:
Obtain the monthly complete vehicle logistics order data of history;
Extract feature construction learning sample set;
According to learning sample set, constructs Random Forest model and be trained and assess, obtain the monthly order of complete vehicle logistics
Prediction model;
The route predicted needed for input predicts that the route is following monthly according to the monthly order forecasting model of complete vehicle logistics
Complete vehicle logistics order simultaneously exports.
In one embodiment, the monthly complete vehicle logistics order data of acquisition history includes:
Data collection steps, the complete vehicle logistics order data monthly to history extract acquisition, form every route
Monthly complete vehicle logistics order time series data;
Data cleansing step, according to preset data cleansing rule, to the monthly complete vehicle logistics order time of acquisition
Sequence is cleaned, and every required route completely monthly complete vehicle logistics order time series data is extracted.
In one embodiment, before data collection steps, further includes:
Data storing steps receive the monthly complete vehicle logistics order data of history and store, form complete vehicle logistics order numbers
According to library.
In one embodiment, the learning sample collection is combined into
(X, Y)={ (xij;t,yij;t), i=1,2 ..., I;J=1,2 ..., J;T=1,2 ..., T }
Wherein, yij;tFor the complete vehicle logistics order volume in t is monthly from departure place i to destination j, xij;tIt is monthly in t
The interior corresponding characteristic vector constructed from departure place i to destination j.
In one embodiment, the characteristic vector xij;tFor the characteristic vector of 13 dimensions, the corresponding inspection of each D feature vectors
Feature is surveyed to be respectively as follows:
Circuit number;
The t monthly corresponding time;
T monthly corresponding month;
T-2 monthly complete vehicle logistics order volume;
The arithmetic mean of instantaneous value of the monthly complete vehicle logistics order volume of t-2, t-3;
The arithmetic mean of instantaneous value of the monthly complete vehicle logistics order of t-2, t-3, t-4;
The weighted average of the monthly complete vehicle logistics order volume of t-2, t-3, wherein t-2 monthly weight 2/3, t-3 is monthly
Weight 1/3;
The weighted average of the monthly complete vehicle logistics order of t-2, t-3, t-4, wherein t-2 monthly weight 3/6, t-3 is monthly
Weight 2/6, t-4 monthly weight 1/6;
The difference of the monthly complete vehicle logistics order of t-2, t-3;
The monthly complete vehicle logistics order minimum value of 1,2 ..., t-2;
The monthly complete vehicle logistics order maximum value of 1,2 ..., t-2;
The monthly complete vehicle logistics orders on average of 1,2 ..., t-2;
The monthly complete vehicle logistics order standard deviation of 1,2 ..., t-2.
In one embodiment, the learning sample set, including for training the training data set of Random Forest model
With the test data set for testing Random Forest model.
In one embodiment, described according to learning sample set, it constructs Random Forest model and is trained and assesses, obtain
The monthly order forecasting model of complete vehicle logistics, comprising:
According to learning sample set, to being trained for the Random Forest model under different hyper parameter set;
The Random Forest model trained is assessed, optimum model parameter set is chosen;
The monthly order forecasting model of complete vehicle logistics is established using the optimum model parameter set by choosing and is trained,
Obtain the monthly order forecasting model of complete vehicle logistics.
In one embodiment, the Random Forest model is established by following steps:
Have and choose several samples in the slave training data set put back to, forms sample set;
Using sample set one decision tree of training, during training decision tree, from characteristic vector when each feature divides
In randomly select several characteristic elements, then therefrom select an optimal feature as disruptive features;
It repeats the above steps, more decision trees of training form random forest;
Each decision tree provides a prediction result, is obtained by voting rule to the forecast sample data of input
The final output of prediction model.
In one embodiment, the Random Forest model obtains the final output of prediction model by voting rule
Specifically, being averaged to the prediction result of each decision tree, the final output of prediction model is obtained.
In one embodiment, by model accuracy formula below, the Random Forest model of building is assessed:
Wherein, y indicates true monthly complete vehicle logistics order volume,Indicate the monthly order of complete vehicle logistics of model prediction
Amount, η is model accuracy.
To achieve the goals above, the present invention also provides a kind of monthly order forecasting device of complete vehicle logistics, feature exists
In, comprising:
Data acquisition module obtains the monthly complete vehicle logistics order data of history;
Feature construction module extracts feature construction learning sample set;
Model training module constructs Random Forest model and is trained and assesses, obtain vehicle according to learning sample set
The monthly order forecasting model of logistics;
Model measurement module is tested the monthly order forecasting model of complete vehicle logistics, and is ordered using complete vehicle logistics are monthly
Single prediction model is predicted and is exported to the following monthly complete vehicle logistics order of the route of required prediction.
In one embodiment, the data acquisition module includes:
Data acquisition module, the complete vehicle logistics order data monthly to history extract acquisition, form every route
Monthly complete vehicle logistics order time series data;
Data cleansing module, according to preset data cleansing rule, to the monthly complete vehicle logistics order time of acquisition
Sequence is cleaned, and every required route completely monthly complete vehicle logistics order time series data is extracted.
In one embodiment, the monthly order forecasting device of the complete vehicle logistics further includes data memory module, receives history
Monthly complete vehicle logistics order data simultaneously stores, and forms complete vehicle logistics order database.
In one embodiment, the learning sample collection of the feature construction module, building is combined into
(X, Y)={ (xij;t,yij;t), i=1,2 ..., I;J=1,2 ..., J;T=1,2 ..., T }
Wherein, yij;tFor the complete vehicle logistics order volume in monthly t from departure place i to destination j, xij;tFor in monthly t
The corresponding characteristic vector constructed from departure place i to destination j.
In one embodiment, the characteristic vector xij;tFor the characteristic vector of 13 dimensions, the corresponding inspection of each D feature vectors
Feature is surveyed to be respectively as follows:
Circuit number;
The t monthly corresponding time;
T monthly corresponding month;
T-2 monthly complete vehicle logistics order volume;
The arithmetic mean of instantaneous value of the monthly complete vehicle logistics order volume of t-2, t-3;
The arithmetic mean of instantaneous value of the monthly complete vehicle logistics order of t-2, t-3, t-4;
The weighted average of the monthly complete vehicle logistics order volume of t-2, t-3, wherein t-2 monthly weight 2/3, t-3 is monthly
Weight 1/3;
The weighted average of the monthly complete vehicle logistics order of t-2, t-3, t-4, wherein t-2 monthly weight 3/6, t-3 is monthly
Weight 2/6, t-4 monthly weight 1/6;
The difference of the monthly complete vehicle logistics order of t-2, t-3;
The monthly complete vehicle logistics order minimum value of 1,2 ..., t-2;
The monthly complete vehicle logistics order maximum value of 1,2 ..., t-2;
The monthly complete vehicle logistics orders on average of 1,2 ..., t-2;
The monthly complete vehicle logistics order standard deviation of 1,2 ..., t-2.
In one embodiment, the learning sample set of the feature construction module, building includes for training random forest
The training data set of model and test data set for testing Random Forest model.
In one embodiment, the model training module includes:
Cross validation module, the learning sample set obtained according to feature construction module, under different hyper parameter set
Random Forest model being trained and assessing;
Parameter chooses module and compares each hyper parameter set, Cong Zhongxuan according to the model performance that cross validation module obtains
Take optimum model parameter set;
It is pre- to establish the monthly order of complete vehicle logistics using the optimum model parameter set by choosing for best model training module
It surveys model and is trained, obtain the monthly order forecasting model of complete vehicle logistics.
In one embodiment, the Random Forest model is established by following steps:
Have and choose several samples in the slave training data set put back to, forms sample set;
Using sample set one decision tree of training, during training decision tree, from characteristic vector when each feature divides
In randomly select several characteristic elements, then therefrom select an optimal feature as disruptive features;
It repeats the above steps, more decision trees of training form random forest;
Each decision tree provides a prediction result, is obtained by voting rule to the forecast sample data of input
The final output of prediction model.
In one embodiment, the Random Forest model obtains the final output of prediction model by voting rule
Specifically, being averaged to the prediction result of each decision tree, the final output of prediction model is obtained.
In one embodiment, the model training module, by model accuracy formula below, to the random gloomy of building
Woods model is assessed:
Wherein, y indicates true monthly complete vehicle logistics order volume,Indicate the monthly order of complete vehicle logistics of model prediction
Amount, η is model accuracy.
To achieve the goals above, the present invention also provides a kind of monthly order forecasting systems of complete vehicle logistics, comprising:
Memory, for storing the instruction that can be executed by processor;
Processor realizes method as described above for executing described instruction.
To achieve the goals above, the present invention also provides a kind of computer storage mediums, are stored thereon with computer and refer to
It enables, wherein executing method as described above when computer instruction is executed by processor.
The monthly order forecast method of complete vehicle logistics provided by the invention and device, system, storage medium, are realized to vehicle object
The prediction of monthly order to be flowed, is the pool transport power plan arrangement of carrier, storehouse management provides more accurately information,
And then reduce production management cost.
The present invention, which specifically has the advantages that, makes full use of artificial intelligence, passes through automation, intelligence, sequencing
Mode predicted, greatly improve the efficiency of prediction, reduce cost of labor, realize relatively high prediction accuracy,
And then reduce the cost of complete vehicle logistics.
Detailed description of the invention
The above and other feature of the present invention, property and advantage will pass through description with reference to the accompanying drawings and examples
And become apparent, identical appended drawing reference always shows identical feature in the accompanying drawings, in which:
Fig. 1 discloses the flow chart of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention;
Fig. 2 discloses the information flow diagram of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention;
The decision tree structure that Fig. 3 a discloses the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention is shown
It is intended to;
Fig. 3 b discloses the feature space signal of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention
Figure;
Fig. 4 discloses the schematic diagram of random forests algorithm according to an embodiment of the invention;
Fig. 5 a discloses the prediction effect figure of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention;
Fig. 5 b discloses the accuracy schematic diagram of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention;
Fig. 6 discloses the structural schematic diagram of the monthly order forecasting device of complete vehicle logistics according to an embodiment of the invention;
Fig. 7 discloses the block diagram of the monthly order forecasting system of complete vehicle logistics of another embodiment according to the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only invented to explain, and do not have to
It is invented in limiting.
The embodiment of the present invention describes the monthly order forecast method of complete vehicle logistics and prediction meanss, system, storage medium.
The executing subject of the monthly order forecast method of complete vehicle logistics provided in an embodiment of the present invention, can be provided in an embodiment of the present invention
The monthly order forecasting device of complete vehicle logistics, system, storage medium, or it is integrated with the monthly order forecasting dress of the complete vehicle logistics
It sets, the terminal device of system, storage medium (for example, smart phone, tablet computer etc.) or server, the system can use hard
Part or software realization.
The present invention complete vehicle logistics order data monthly by acquisition history, and the cycle information of order is excavated, history wave
Dynamic model formula, spatial information construct the characteristic of the monthly order of complete vehicle logistics, and pass through the random forests algorithm in machine learning
The mapping relations of order feature and order to be predicted are established, and then are realized to the monthly order forecasting of complete vehicle logistics.
Refering to what is shown in Fig. 1, Fig. 1 discloses the stream of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention
Cheng Tu.
The monthly order forecast method of the complete vehicle logistics, mainly comprises the following steps that
S101, the monthly complete vehicle logistics order data of history is obtained;
S102, feature construction learning sample set is extracted;
S103, according to learning sample set, construct Random Forest model and be trained and assess, it is monthly to obtain complete vehicle logistics
Order forecasting model;
The route of prediction needed for S104, input predicts the route following moon according to the monthly order forecasting model of complete vehicle logistics
The complete vehicle logistics order of degree simultaneously exports.
Fig. 2 discloses the information flow diagram of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention.
Referring to fig. 1 and fig. 2, detailed expansion is carried out to each step in the monthly order forecast method of complete vehicle logistics below to explain
It states.
S101, the monthly complete vehicle logistics order data of history is obtained.
In order to predict the monthly order of complete vehicle logistics, need to obtain the monthly complete vehicle logistics order data of history, so
The following monthly complete vehicle logistics order data of prediction model prediction is passed through based on the monthly complete vehicle logistics order data of history afterwards.
In one embodiment, obtaining the monthly complete vehicle logistics order data of history includes:
S202, data collection steps, the complete vehicle logistics order data monthly to history are carried out according to the requirement of prediction model
Acquisition is extracted, the monthly complete vehicle logistics order time series data of every route is formed.
S203, data cleansing step, according to preset data cleansing rule, to the monthly complete vehicle logistics order of acquisition
Time series is cleaned, and every required route completely monthly complete vehicle logistics order time series data is extracted.It is described pre-
The data cleansing rule first set, the including but not limited to historical data of removal part exception and some missings of completion are gone through
History data.
By data cleansing step S203, the exceptional value in History Order data is removed, exceptional value is avoided to influence
The accuracy of subsequent prediction carries out processing completion to missing values, improves the accuracy of subsequent prediction.
In specific implementation, the monthly complete vehicle logistics order data of the history of acquisition may not be able to directly meet prediction mould
The requirement of type.Therefore it needs to pre-process the monthly complete vehicle logistics order data of the history of acquisition, such as data acquire sum number
According to cleaning, to obtain the monthly complete vehicle logistics order time series data for meeting prediction model requirement.
In another embodiment of the present invention, before data collection steps S202 further include:
S201, data storing steps receive the monthly complete vehicle logistics order data of history and store, and form complete vehicle logistics and order
Single database.Complete vehicle logistics order data is stored in the form of database, acquires work convenient for management and subsequent data
Make.Data collection steps S202 realizes the complete vehicle logistics order data monthly to history according to prediction model by database interface
Requirement extract acquisition.
Further, the monthly complete vehicle logistics order data of history is inputted order management by data storing steps S201
System is stored, and complete vehicle logistics order database is formed.Data collection steps S202 is realized by database interface to order
The monthly complete vehicle logistics order data of history in management system extracts acquisition, forms the monthly complete vehicle logistics of every route
Order time series data.
It is understood that in order to make acquired complete vehicle logistics order data meet the requirement of prediction model, avoid pair
The influence of subsequent prediction can carry out the pretreatment of other forms to order data, can also be according to acquired order data
Concrete form, what progress was selective as requested pre-process, and such as only carries out a data collection steps, without data cleansing step
Rapid etc., details are not described herein again.If acquired order data has complied fully with the requirement of prediction model, can also will be obtained
The complete vehicle logistics order data taken is directly entered the operation of next step without being pre-processed.
S102, feature construction learning sample set is extracted.
According to acquired complete vehicle logistics order data, by excavating historical information, spatial information constructs learning sample collection
It closes, is supplied to prediction model and carries out learning training.
In the embodiment shown in Figure 2, which is specially S204 feature construction step.For the convenience of description, yij;tFor
In the complete vehicle logistics order volume of the monthly interior j from origin to destination of t.
According to the monthly complete vehicle logistics order data of history that step S201, S202, S203 are obtained, the monthly order time is obtained
Sequences yij;t, wherein i=1,2 ..., I, j=1,2 ..., J, t=1,2 ..., T, that is, there are I departure place and J destination,
The monthly complete vehicle logistics order data of T monthly history.According to the monthly available order time series data before of t
yij;1, yij;2..., yij;t-2Predict the monthly monthly complete vehicle logistics order of t.
xij;tFor the corresponding characteristic vector constructed in t is monthly from departure place i to destination j.xij;tFor the arrow of 13 dimensions
Amount, xij;tIt (n) is xij;tNth elements.
The meaning of each characteristic element in characteristic vector is illustrated below.
xij;t(1): route ID takes the only integer in 1~L, and wherein L is route sum;
xij;t(2): the t monthly corresponding time is 2012,2013 ..., 2017 according to current historical data value;
xij;t(3): t monthly corresponding month is 1,2 ..., 12 according to current historical data value;
xij;t(4): t-2 monthly complete vehicle logistics order volume yij;t-2;
xij;t(5): the arithmetic mean of instantaneous value of the monthly complete vehicle logistics order volume of t-2, t-3
xij;t(6): the arithmetic mean of instantaneous value of the monthly complete vehicle logistics order of t-2, t-3, t-4
xij;t(7): the weighted average of the monthly complete vehicle logistics order volume of t-2, t-3
xij;t(8): the weighted average of the monthly complete vehicle logistics order of t-2, t-3, t-4
xij;t(9): the difference y of the monthly complete vehicle logistics order of t-2, t-3ij;t-2-yij;t-3;
xij;t(10): the monthly complete vehicle logistics order minimum value x of 1,2 ..., t-2ij;t(10)=min { yij;1,yij;2,...,
yij;t-2};
xij;t(11): the monthly complete vehicle logistics order maximum value x of 1,2 ..., t-2ij;t(11)=max { yij;1,yij;2,...,
yij;t-2};
xij;t(12): the monthly complete vehicle logistics orders on average of 1,2 ..., t-2
xij;t(13): the monthly complete vehicle logistics order standard deviation of 1,2 ..., t-2
By above feature construction method, following learning sample set is generated:
(X, Y)={ (xij;t,yij;t), i=1,2 ..., I;J=1,2 ..., J;T=1,2 ..., T }.
In order to which subsequent algorithm describes conveniently, learning sample to be denoted as (x againm,ym), m=1,2 ..., M, wherein M is sample
This number, and remember learning sample set ξ={ (xm,ym), m=1,2 ..., M }.
Further, by learning sample set ξ={ (xm,ym), m=1,2 ..., M } it is divided into training set ψ={ (xm,ym),
M=1,2 ..., p } and test set ζ={ (xm,ym), m=p+1, p+2 ..., M }, wherein p is the number of samples of training set.
Training set ψ is for training Random Forest model, and test set ζ is for testing Random Forest model.
S103, according to learning sample set, construct Random Forest model and be trained and assess, it is monthly to obtain complete vehicle logistics
Order forecasting model.
According to the learning sample set that step S102 is constructed, the training monthly order forecasting model of complete vehicle logistics, to what is obtained
Prediction model is assessed, if assessment passes through, completes the building of the monthly order forecasting model of complete vehicle logistics.
In the embodiment shown in Figure 2, step S103 specifically include the following three steps:
The first step, according to the data in learning sample set, to the Random Forest model under different hyper parameter set into
Row training.In the embodiment shown in Figure 2, which is S205 cross validation step.
Hyper parameter refers to the parameter of prediction model required setting before starting machine-learning process.As an example, pre-
The hyper parameter for surveying model includes but is not limited to the quantity set, the depth of tree, the number of leaf node.Other ginsengs other than hyper parameter
Number needs to obtain by training.
The setting of hyper parameter has large effect to the performance of prediction model, it is therefore desirable to optimize, select to hyper parameter
One group of optimal hyper parameter is selected, to improve the performance and effect of study.
Optionally, by the method for cross validation, it is different that prediction model is carried out to multiple groups under different hyper parameter set
Training obtains the prediction model of multiple groups difference hyper parameter set.In the embodiment shown in Figure 2, it by cross validation method, obtains
Obtain n model parameter set, i.e. n different hyper parameter set.
The mode of the cross validation carries out obtained learning sample data more to reuse learning sample data
Secondary cutting, group are combined into different training set and test set, train prediction model with training set, with test set come assessment prediction mould
The quality of type.Available multiple groups different training set and test set on this basis, certain sample in certain training set is under
The secondary sample being likely to become in test set.
Second step assesses the Random Forest model under the multiple groups difference hyper parameter set trained, chooses best
Model parameter set.
In the embodiment shown in Figure 2, which is that S206 model compares step, and S205 cross validation step is obtained
Multiple prediction models carry out Performance Evaluation, choose optimum model parameter set.
The mode of prediction model Performance Evaluation can be a variety of.In the embodiment shown in Figure 2, pass through the accurate of model
Degree assesses the random forest of building.Model accuracy formula is as follows:
Wherein, y indicates true monthly complete vehicle logistics order volume,Indicate the monthly order of complete vehicle logistics of model prediction
Amount, η is model accuracy.
Third step establishes the monthly order forecasting model of complete vehicle logistics and goes forward side by side using the optimum model parameter set by choosing
Row training, obtains the monthly order forecasting model of complete vehicle logistics.
In the embodiment shown in Figure 2, which is S207 best model training step, optimal due to choosing in the first step
During hyper parameter, it is trained using the data of the partial data in learning sample set, i.e. training set, test set
There is no use in training for data.Therefore after optimal hyper parameter has been determined, using whole learning sample data, i.e., in training
It is trained again on the basis of the sum of data of collection and test set, finally obtains the monthly order forecasting model of complete vehicle logistics.
The present invention constructs the monthly order forecasting model of complete vehicle logistics using random forest (Random Forest, RF) algorithm.
It is illustrated below to how random forests algorithm constructs the monthly order forecasting model of complete vehicle logistics.
Random forests algorithm refers to setting a kind of algorithm for being trained sample and predicting using more.It is following go out
This algorithm is described in detail in version object, " The Elements of Statistical Learning ", author
Trevor Hastie/Robert Tibshirani/Jerome Friedman。
The training step of Random Forest model is as follows in the present invention:
Have and choose several samples in the slave training data set put back to, forms sample set;
Using sample set one decision tree of training, during training decision tree, from characteristic vector when each feature divides
In randomly select several characteristic elements, then therefrom select an optimal feature as disruptive features;
It repeats the above steps, more decision trees of training form random forest;
Each decision tree provides a prediction result, is obtained by voting rule to the forecast sample data of input
The final output of prediction model.
In the embodiment shown in Figure 2, the prediction result of each decision tree is averaged, obtains prediction model most
Output result eventually.
Below in conjunction with formula, the training process of the order forecasting model based on random forests algorithm is elaborated.
1) training set data ψ={ (x is inputtedm,ym), m=1,2 ..., p }.
2) to the number of iterations b=1,2 ..., B,
2.1) there is slave the training dataset ψ={ (x put back tom,ym), m=1,2 ..., p } n sample of selection;
2.2) one decision tree T (x of sample set training of sampling is utilized;Θb), during training decision tree, every time
M characteristic element first is selected from the M all characteristic elements of characteristic vector at random when feature divides, then from this m spy
Select an optimal feature as disruptive features in sign;In general, m should be much smaller than M, in the growth course of entire forest
In, m is remained unchanged.In the embodiment shown in Figure 2,13 M, m 4.
3) final output random forest
In above formula, it is also the number of iterations that B, which is the tree of decision tree,.B value can be manually set, after training to setting number
That is deconditioning.
ΘbFor the parameter sets of decision tree.
Further, ΘbCut-off comprising characteristic element and leaf weight cn。
To which decision tree can be further represented as
Rn is corresponding leaf node subregion, and n=1 ..., N, N are leaf
The number of node, I are indicator function.
In one embodiment, leaf weight cn=ave (ym|x∈Rn), i.e., leaf weight is to belong to the leaf node subregion
The mean value of the label value of all learning samples.
In order to further explain the structure of decision tree, single is illustrated for the case where 2 dimension with characteristic vector
The structure of decision tree, as shown in Figure 3a and Figure 3b shows, Fig. 3 a disclose the monthly order of complete vehicle logistics according to an embodiment of the invention
The decision tree structure schematic diagram of prediction technique, Fig. 3 b disclose feature space schematic diagram.The characteristic element of characteristic vector is X1With
X2, the cut-off of characteristic element is t1~t4, determine cut-off, constitute decision tree T (x as shown in Figure 3a;Θb) structure, will
Feature space is divided into R as shown in Figure 3b1~R5Five leaf node subregions.
Fig. 4 discloses the schematic diagram of random forests algorithm according to an embodiment of the invention.In order to facilitate signal, with three
Random forests algorithm is illustrated for decision tree.Former training dataset 41, the training set of every decision tree are instructed to original
Practice after data set 41 carries out the sampling put back to and obtains.The first instruction is obtained after carrying out the sampling put back to former training dataset 41
Practice data set 411, the second training dataset 411, third training dataset 413, is obtained using the training of the first training dataset 411
First decision tree 421 obtains the second decision tree 422 using the training of the second training dataset 412, utilizes third training dataset
412 training obtain third decision tree 423, and final Random Forest model 42 is the mean value of all decision tree fitting results.At random
The advantages of forest algorithm is that generalization ability is strong, the disadvantage is that fitting precision is poor.
The route of prediction needed for S104, input predicts the route following moon according to the monthly order forecasting model of complete vehicle logistics
The complete vehicle logistics order of degree simultaneously exports.
In the embodiment shown in Figure 2, which is S208 model measurement step, the complete vehicle logistics moon that final training is completed
Order forecasting model is spent, all routes are predicted, calculates the complete vehicle logistics order of following every monthly route.
It is defeated according to final mask prediction steps S209 when the monthly order of the complete vehicle logistics for needing to predict a certain route
Enter this route, according to the operation result of the monthly order forecasting model of complete vehicle logistics, returns to the prediction of the monthly order of complete vehicle logistics
As a result, and being exported by output equipment.
To better understand those skilled in the art and implementing the present invention, the embodiment of the present invention gives the reference present invention
The contrast effect figure of scheme and true monthly complete vehicle logistics order volume that embodiment provides.
Fig. 5 a discloses the prediction effect figure of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention, figure
5b discloses the accuracy schematic diagram of the monthly order forecast method of complete vehicle logistics according to an embodiment of the invention.
Use the monthly logistics order data of the vehicle of in March, 2012 in August, 2017 for training set, and will be by above-mentioned whole
The monthly order forecast method of vehicle logistics, in test set, i.e., the complete vehicle logistics of in Septembers, 2017 are monthly to be ordered the prediction model that training is completed
It is tested in forms data.
By taking this stream line of Shanghai to Shanghai as an example, prediction model is as shown in Figure 5 a in the effect that training set closes.Its
Middle abscissa indicates monthly, and ordinate indicates complete vehicle logistics order volume, and y curve represents true complete vehicle logistics order volume,It is bent
Line indicates the complete vehicle logistics order volume that model is predicted on training set.
It vivid can find out that prediction model compares accurate fitting to the order data of training set from Fig. 5 a.
For the effect of further quantitative model, by model accuracy formula below, to the random forest mould of building
Type is assessed:
Wherein, y indicates true monthly complete vehicle logistics order volume,Indicate the monthly order of complete vehicle logistics of model prediction
Amount, η is model accuracy.
Fig. 5 b illustrates performance of the Shanghai to Shanghai route complete vehicle logistics order model accuracy on training set.Due to reality
Order data statistics in 2012 and the data statistics in other times later have differences in border, lead to table of the model in 2012
Performance annual now and later is inconsistent.
But Fig. 5 b can be seen that in the case where order data counts unanimous circumstances, prediction model reaches on training set
Very high prediction accuracy.
The monthly order forecasting forecast result of model of real assessment complete vehicle logistics, will see table of the prediction model on test set
Existing, the performance by taking Shanghai to Shanghai route as an example to prediction model on test set is illustrated.
To the practical complete vehicle logistics order volume y=5944 of Shanghai route, the complete vehicle logistics of model prediction are ordered in the Shanghai of in September, 2017
List is measuredModel accuracy η=0.912.As it can be seen that prediction model has similarly reached very high on test set
Prediction accuracy.
It orders the present invention also provides a kind of complete vehicle logistics that the monthly order forecast method of above-mentioned complete vehicle logistics may be implemented are monthly
Single prediction device.As shown in fig. 6, the monthly order forecasting device of the complete vehicle logistics includes data acquisition module 61, feature construction mould
Block 62, model training module 63 and model measurement module 64.
Data acquisition module 61, output end are connect with feature construction module 62, including data acquisition module 611 and data it is clear
Mold cleaning block 612.
Wherein, data acquisition module 611, output end are connect with data cleansing module 612, the complete vehicle logistics monthly to history
Order data extracts acquisition, forms the monthly complete vehicle logistics order time series data of every route.
Data cleansing module 612, output end are connect with feature construction module 62, are advised according to preset data cleansing
Then, the monthly complete vehicle logistics order time series of acquisition is cleaned, extracts every required route completely monthly vehicle
Logistics order time series data.The preset data cleansing rule, including but not limited to removal part is abnormal goes through
The historical data of history data and some missings of completion.
Further, the monthly order forecasting device of complete vehicle logistics further includes data memory module 65, output end and data
It obtains module 61 to connect, receive the monthly complete vehicle logistics order data of history and stores, form complete vehicle logistics order database.Number
It according to the data acquisition module 611 for obtaining module 61, connect with data memory module 65, data is deposited by database interface realization
The acquisition for storing up data in module 65 is extracted.
In one embodiment, data memory module 65 is order management system, and data acquisition module 611 passes through database
Interface realizes that the acquisition to data in order management system is extracted.
Feature construction module 62, output end are connect with model training module 63, by excavate historical information, spatial information,
Extract feature construction learning sample set.
The learning sample collection that feature construction module 62 constructs is combined into
(X, Y)={ (xij;t,yij;t), i=1,2 ..., I;J=1,2 ..., J;T=1,2 ..., T }
Wherein, yij;tFor the complete vehicle logistics order volume in monthly t from departure place i to destination j, xij;tFor in monthly t
The corresponding characteristic vector constructed from departure place i to destination j.
Monthly order time series yij;t, wherein i=1,2 ..., I, j=1,2 ..., J, t=1,2 ..., T, that is, there is I
A departure place and J destination, the monthly complete vehicle logistics order data of T monthly history.It is monthly available before according to t
Order time series data yij;1, yij;2..., yij;t-2Predict the monthly monthly complete vehicle logistics order of t.
xij;tFor the vector of 13 dimensions, xij;tIt (n) is xij;tNth elements.For each characteristic element in characteristic vector
Meaning be described in detail in the embodiment of the monthly order forecast method of above-mentioned complete vehicle logistics, which is not described herein again.
According to above feature construction method, feature construction module generates learning sample set:
(X, Y)={ (xij;t,yij;t), i=1,2 ..., I;J=1,2 ..., J;T=1,2 ..., T }.
Learning sample is denoted as (x againm,ym), m=1,2 ..., M, wherein M is number of samples, and remembers learning sample collection
Close ξ={ (xm,ym), m=1,2 ..., M }.
Further, by learning sample set ξ={ (xm,ym), m=1,2 ..., M } it is divided into training set ψ={ (xm,ym),
M=1,2 ..., p } and test set ζ={ (xm,ym), m=p+1, p+2 ..., M }, wherein p is the number of samples of training set.
Training set ψ is for training Random Forest model, and test set ζ is for testing Random Forest model.
Model training module 63, output end are connect with model measurement module 64, including the choosing of cross validation module 631, parameter
Modulus block 632 and best model training module 633.
Cross validation module 631, input terminal are connect with feature construction module 62, and output end and parameter are chosen module 632 and connected
It connects, the learning sample set obtained according to feature construction module instructs the Random Forest model under different hyper parameter set
Practice and assesses.
Parameter chooses module 632, and output end connect with best model training module 633, obtained according to cross validation module
Model performance, further compare parameters set, and therefrom choose optimum model parameter set.
Best model training module 633, output end are connect with model measurement module 64, utilize the best model by choosing
Parameter sets are established the monthly order forecasting model of complete vehicle logistics and are trained in full dose data, obtain that complete vehicle logistics are monthly to be ordered
Single prediction model.
Model measurement module 64, input terminal are connect with model training module 63, to the monthly order forecasting model of complete vehicle logistics
It is tested, and utilizes the following monthly complete vehicle logistics order of route of the monthly order forecasting model of complete vehicle logistics to required prediction
It is predicted and is exported.
In one embodiment, model measurement module 64 predicts all routes, calculates following monthly every
The complete vehicle logistics order of route.When the monthly order of the complete vehicle logistics for needing to predict a certain route, by input equipment (such as visitor
Family end) this route is inputted, model measurement module 64 returns whole according to the operation result of the monthly order forecasting model of complete vehicle logistics
The prediction result of the monthly order of vehicle logistics, and exported by output equipment.
Fig. 7 is the block diagram of the monthly order forecasting system of complete vehicle logistics of further embodiment of this invention.Complete vehicle logistics are monthly to be ordered
Single prediction system may include internal communication bus 701, processor (processor) 702, read-only memory (ROM) 703, random
Access memory (RAM) 704, communication port 705 and hard disk 707.The complete vehicle logistics moon may be implemented in internal communication bus 701
Spend the data communication between order forecasting system component.Processor 702 can be judged and be issued prompt.In some embodiments
In, processor 702 can be made of one or more processors.
Communication port 705 may be implemented between the monthly order forecasting system of complete vehicle logistics and the input-output apparatus of outside
Carry out data transmission and communicates.In some embodiments, the monthly order forecasting system of complete vehicle logistics can pass through communication port 705
It is sent and received information from network and data.In some embodiments, the monthly order forecasting system of complete vehicle logistics can be by defeated
Enter/output end 706 to be to carry out data transmission and communicate between wired form and the input-output apparatus of outside.
The monthly order forecasting system of complete vehicle logistics can also include various forms of program storage units and data storage
Unit, such as hard disk 707, read-only memory (ROM) 703 and random access memory (RAM) 704, can store at computer
Possible program instruction performed by the various data files and processor 702 that reason and/or communication use.Processor 702 is held
These instructions of row are with the major part of implementation method.The result of the processing of processor 702 is transmitted to the defeated of outside by communication port 705
Equipment out is shown in the user interface of output equipment.
For example, the implementation process file of the monthly order forecasting method of above-mentioned complete vehicle logistics can be computer program,
It is stored in hard disk 707, and can be recorded in processor 702 and execute, to implement the present processes.
When the implementation process file of the monthly order forecast method of complete vehicle logistics is computer program, it also can store and calculating
Product is used as in machine readable storage medium storing program for executing.For example, computer readable storage medium can include but is not limited to magnetic storage apparatus (example
Such as, hard disk, floppy disk, magnetic stripe), CD (for example, compact disk (CD), digital versatile disc (DVD)), smart card and flash memory device
(for example, electrically erasable programmable read-only memory (EPROM), card, stick, key drive).In addition, various storages described herein are situated between
Mass-energy represents one or more equipment and/or other machine readable medias for storing information.Term " machine readable media "
Can include but is not limited to store, include and/or the wireless channel of carrying code and/or instruction and/or data and it is various its
Its medium (and/or storage medium).
The monthly order forecast method of complete vehicle logistics provided by the invention and device, system, storage medium, are realized to vehicle object
The prediction of monthly order to be flowed, is the pool transport power plan arrangement of carrier, storehouse management provides more accurately information,
And then reduce production management cost.
The present invention, which specifically has the advantages that, makes full use of artificial intelligence, passes through automation, intelligence, sequencing
Mode predicted, greatly improve the efficiency of prediction, reduce cost of labor, realize relatively high prediction accuracy,
And then reduce the cost of complete vehicle logistics.
Although for simplify explain the above method is illustrated to and is described as a series of actions, it should be understood that and understand,
The order that these methods are not acted is limited, because according to one or more embodiments, some movements can occur in different order
And/or with from it is depicted and described herein or herein it is not shown and describe but it will be appreciated by those skilled in the art that other
Movement concomitantly occurs.
As shown in the application and claims, unless context clearly prompts exceptional situation, " one ", "one", " one
The words such as kind " and/or "the" not refer in particular to odd number, may also comprise plural number.It is, in general, that term " includes " only prompts to wrap with "comprising"
Include clearly identify the step of and element, and these steps and element do not constitute one it is exclusive enumerate, method or apparatus
The step of may also including other or element.
It should be understood that embodiments described above is only signal.Embodiment described herein can be in hardware, software, solid
It is realized in part, middleware, microcode or any combination thereof.For hardware realization, processing unit can be in one or more spy
Determine purposes integrated circuit (ASIC), digital signal processor (DSP), digital signal processing appts (DSPD), programmable logic device
It part (PLD), field programmable gate array (FPGA), processor, controller, microcontroller, microprocessor and/or is designed as executing
Other electronic units of function described herein or its combine interior realization.
Above-described embodiment, which is available to, to be familiar with person in the art to realize or use the present invention, and is familiar with this field
Personnel can make various modifications or variation, thus this to above-described embodiment without departing from the present invention in the case of the inventive idea
The protection scope of invention is not limited by above-described embodiment, and should meet inventive features that claims are mentioned most
On a large scale.
Claims (10)
1. a kind of complete vehicle logistics order forecast method, which comprises the following steps:
Obtain the monthly complete vehicle logistics order data of history;
Extract feature construction learning sample set;
According to learning sample set, constructs Random Forest model and be trained and assess, obtain the monthly order forecasting of complete vehicle logistics
Model;
The route predicted needed for input predicts the following monthly vehicle of the route according to the monthly order forecasting model of complete vehicle logistics
Logistics order simultaneously exports.
2. complete vehicle logistics order forecast method according to claim 1, which is characterized in that the monthly vehicle of the acquisition history
Logistics order data includes:
Data collection steps, the complete vehicle logistics order data monthly to history extract acquisition, form the monthly of every route
Complete vehicle logistics order time series data;
Data cleansing step, according to preset data cleansing rule, to the monthly complete vehicle logistics order time series of acquisition
It is cleaned, extracts every required route completely monthly complete vehicle logistics order time series data.
3. complete vehicle logistics order forecast method according to claim 1, which is characterized in that the learning sample collection is combined into
(X, Y)={ (xij;t,yij;t), i=1,2 ..., I;J=1,2 ..., J;T=1,2 ..., T }
Wherein, yij;tFor the complete vehicle logistics order volume in t is monthly from departure place i to destination j, xij;tIt is monthly interior from out in t
The corresponding characteristic vector that hair ground i to destination j is constructed.
4. complete vehicle logistics order forecast method according to claim 3, which is characterized in that the characteristic vector xij;tIt is 13
The characteristic vector of dimension, the corresponding detection feature of each D feature vectors are respectively as follows:
Circuit number;
The t monthly corresponding time;
T monthly corresponding month;
T-2 monthly complete vehicle logistics order volume;
The arithmetic mean of instantaneous value of the monthly complete vehicle logistics order volume of t-2, t-3;
The arithmetic mean of instantaneous value of the monthly complete vehicle logistics order of t-2, t-3, t-4;
The weighted average of the monthly complete vehicle logistics order volume of t-2, t-3, wherein the monthly weight 2/3 of t-2, t-3 monthly weight
1/3;
The weighted average of the monthly complete vehicle logistics order of t-2, t-3, t-4, wherein the monthly weight 3/6 of t-2, t-3 monthly power
Weigh the monthly weight 1/6 of 2/6, t-4;
The difference of the monthly complete vehicle logistics order of t-2, t-3;
The monthly complete vehicle logistics order minimum value of 1,2 ..., t-2;
The monthly complete vehicle logistics order maximum value of 1,2 ..., t-2;
The monthly complete vehicle logistics orders on average of 1,2 ..., t-2;
The monthly complete vehicle logistics order standard deviation of 1,2 ..., t-2.
5. complete vehicle logistics order forecast method according to claim 3, which is characterized in that the learning sample set, packet
Include the training data set for training Random Forest model and the test data set for testing Random Forest model.
6. complete vehicle logistics order forecast method according to claim 5, which is characterized in that described according to learning sample collection
It closes, building Random Forest model is trained and assesses, and obtains the monthly order forecasting model of complete vehicle logistics, comprising:
According to learning sample set, to being trained for the Random Forest model under different hyper parameter set;
The Random Forest model trained is assessed, optimum model parameter set is chosen;
The monthly order forecasting model of complete vehicle logistics is established using the optimum model parameter set by choosing and is trained, and is obtained
The monthly order forecasting model of complete vehicle logistics.
7. complete vehicle logistics order forecast method according to claim 6, which is characterized in that the Random Forest model passes through
Following steps are established:
Have and choose several samples in the slave training data set put back to, forms sample set;
Using sample set training one decision tree, training decision tree during, each feature divide when from characteristic vector with
Machine chooses several characteristic elements, then therefrom selects an optimal feature as disruptive features;
It repeats the above steps, more decision trees of training form random forest;
Each decision tree provides a prediction result, is predicted by voting rule to the forecast sample data of input
The final output of model.
8. a kind of complete vehicle logistics order forecasting device characterized by comprising
Data acquisition module obtains the monthly complete vehicle logistics order data of history;
Feature construction module extracts feature construction learning sample set;
Model training module constructs Random Forest model and is trained and assesses, obtain complete vehicle logistics according to learning sample set
Monthly order forecasting model;
Model measurement module tests the monthly order forecasting model of complete vehicle logistics, and pre- using the monthly order of complete vehicle logistics
It surveys model and the following monthly complete vehicle logistics order of the route of required prediction is predicted and exported.
9. a kind of logistics system, comprising:
Memory, for storing the instruction that can be executed by processor;
Processor, for executing described instruction to realize the method according to claim 1 to 7.
10. a kind of computer-readable medium, is stored thereon with computer instruction, wherein when computer instruction is executed by processor
When, execute the method according to claim 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910382944.2A CN110309948A (en) | 2019-05-09 | 2019-05-09 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910382944.2A CN110309948A (en) | 2019-05-09 | 2019-05-09 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110309948A true CN110309948A (en) | 2019-10-08 |
Family
ID=68074676
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910382944.2A Pending CN110309948A (en) | 2019-05-09 | 2019-05-09 | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110309948A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111915100A (en) * | 2020-08-17 | 2020-11-10 | 北京交通大学 | High-precision freight prediction method and freight prediction system |
CN113516173A (en) * | 2021-05-27 | 2021-10-19 | 江西五十铃汽车有限公司 | Evaluation method for static and dynamic interference of whole vehicle based on random forest and decision tree |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016025357A2 (en) * | 2014-08-13 | 2016-02-18 | Microsoft Technology Licensing, Llc | Distributed stage-wise parallel machine learning |
CN108305095A (en) * | 2017-12-22 | 2018-07-20 | 杭州先临三维云打印技术有限公司 | Three dimensional model printing Cost Estimation and device |
US20180260531A1 (en) * | 2017-03-10 | 2018-09-13 | Microsoft Technology Licensing, Llc | Training random decision trees for sensor data processing |
US20180260719A1 (en) * | 2017-03-10 | 2018-09-13 | Microsoft Technology Licensing, Llc | Cascaded random decision trees using clusters |
CN108564326A (en) * | 2018-04-19 | 2018-09-21 | 安吉汽车物流股份有限公司 | Prediction technique and device, computer-readable medium, the logistics system of order |
CN108960431A (en) * | 2017-05-25 | 2018-12-07 | 北京嘀嘀无限科技发展有限公司 | The prediction of index, the training method of model and device |
CN108960537A (en) * | 2018-08-17 | 2018-12-07 | 安吉汽车物流股份有限公司 | The prediction technique and device of logistics order, readable medium |
CN109242105A (en) * | 2018-08-17 | 2019-01-18 | 第四范式(北京)技术有限公司 | Tuning method, apparatus, equipment and the medium of hyper parameter in machine learning model |
CN109544196A (en) * | 2017-09-21 | 2019-03-29 | 北京嘀嘀无限科技发展有限公司 | The price multiple of reservation order determines method and apparatus |
-
2019
- 2019-05-09 CN CN201910382944.2A patent/CN110309948A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016025357A2 (en) * | 2014-08-13 | 2016-02-18 | Microsoft Technology Licensing, Llc | Distributed stage-wise parallel machine learning |
US20180260531A1 (en) * | 2017-03-10 | 2018-09-13 | Microsoft Technology Licensing, Llc | Training random decision trees for sensor data processing |
US20180260719A1 (en) * | 2017-03-10 | 2018-09-13 | Microsoft Technology Licensing, Llc | Cascaded random decision trees using clusters |
CN108960431A (en) * | 2017-05-25 | 2018-12-07 | 北京嘀嘀无限科技发展有限公司 | The prediction of index, the training method of model and device |
CN109544196A (en) * | 2017-09-21 | 2019-03-29 | 北京嘀嘀无限科技发展有限公司 | The price multiple of reservation order determines method and apparatus |
CN108305095A (en) * | 2017-12-22 | 2018-07-20 | 杭州先临三维云打印技术有限公司 | Three dimensional model printing Cost Estimation and device |
CN108564326A (en) * | 2018-04-19 | 2018-09-21 | 安吉汽车物流股份有限公司 | Prediction technique and device, computer-readable medium, the logistics system of order |
CN108960537A (en) * | 2018-08-17 | 2018-12-07 | 安吉汽车物流股份有限公司 | The prediction technique and device of logistics order, readable medium |
CN109242105A (en) * | 2018-08-17 | 2019-01-18 | 第四范式(北京)技术有限公司 | Tuning method, apparatus, equipment and the medium of hyper parameter in machine learning model |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111915100A (en) * | 2020-08-17 | 2020-11-10 | 北京交通大学 | High-precision freight prediction method and freight prediction system |
CN111915100B (en) * | 2020-08-17 | 2024-04-05 | 北京交通大学 | High-precision freight prediction method and freight prediction system |
CN113516173A (en) * | 2021-05-27 | 2021-10-19 | 江西五十铃汽车有限公司 | Evaluation method for static and dynamic interference of whole vehicle based on random forest and decision tree |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106549772A (en) | Resource prediction method, system and capacity management device | |
CN110362612A (en) | Abnormal deviation data examination method, device and the electronic equipment executed by electronic equipment | |
CN102073708A (en) | Large-scale uncertain graph database-oriented subgraph query method | |
CN110309947A (en) | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium | |
CN113268641B (en) | User data processing method based on big data and big data server | |
CN110706015A (en) | Advertisement click rate prediction oriented feature selection method | |
CN110309948A (en) | Complete vehicle logistics order forecast method and device, logistics system and computer-readable medium | |
CN106648839A (en) | Method and device for processing data | |
CN115114484A (en) | Abnormal event detection method and device, computer equipment and storage medium | |
CN111325255B (en) | Specific crowd delineating method and device, electronic equipment and storage medium | |
CN107908696A (en) | A kind of parallel efficiently multidimensional space data clustering algorithm GRIDEN based on grid and density | |
CN109767263A (en) | Business revenue data predication method, device, computer equipment and storage medium | |
CN110059880B (en) | Service discovery method and device | |
CN115801557A (en) | Fault root cause positioning method and device and readable storage medium | |
Wirawan et al. | Application of data mining to prediction of timeliness graduation of students (a case study) | |
CN103942403A (en) | Method and device for screening mass variables | |
Park et al. | Application of gap-constraints given sequential frequent pattern mining for protein function prediction | |
CN107995643A (en) | Network element parameter prediction optimization method and device based on data-driven | |
CN109284861A (en) | Prediction processing method, medium, device and calculating equipment | |
CN117435870B (en) | Load data real-time filling method, system, equipment and medium | |
KR101374379B1 (en) | Apparatus and Graph Classification using Co-occurrent Frequent Subgraphs | |
CN118051626A (en) | Business recommendation method and device based on knowledge graph and electronic equipment | |
Long et al. | Machine Learning-Based Predictive Models for Energy Consumption Estimation in Energy-Efficient Building Envelope Design | |
JP2024516440A (en) | Systems and methods for data classification | |
CN113392226A (en) | Method, device and equipment for determining risk entity |
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 |