CN110110894A - Construction method, device, medium, the electronic equipment of Economic Forecasting Mathematical Model - Google Patents
Construction method, device, medium, the electronic equipment of Economic Forecasting Mathematical Model Download PDFInfo
- Publication number
- CN110110894A CN110110894A CN201910273964.6A CN201910273964A CN110110894A CN 110110894 A CN110110894 A CN 110110894A CN 201910273964 A CN201910273964 A CN 201910273964A CN 110110894 A CN110110894 A CN 110110894A
- Authority
- CN
- China
- Prior art keywords
- prediction
- mathematical model
- node
- training sample
- economic forecasting
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
-
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Human Resources & Organizations (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Technology Law (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to electronic digit technical field of data processing, disclose construction method, device, medium and the electronic equipment of a kind of Economic Forecasting Mathematical Model.This method comprises: building Economic Forecasting Mathematical Model frame;Initialize the parameter of each node in the Economic Forecasting Mathematical Model frame;The training sample set for constructing Economic Forecasting Mathematical Model is obtained, includes the desired probability distribution of several sample characteristics and the corresponding prediction result of training sample in the training sample;Based on the training sample in the training sample set, the prior initialized Economic Forecasting Mathematical Model frame of training, to obtain Economic Forecasting Mathematical Model.It is capable of the parameter scale of compact model by the method, and balances the generalization and interpretation of Classical forecast model, the interpretation of model can be improved, so as to makes the objective economy of model explanation and business decision regular.
Description
Technical field
The present invention relates to electronic digit technical field of data processing, in particular to a kind of building side of Economic Forecasting Mathematical Model
Method, device, medium and electronic equipment.
Background technique
Currently, trading, in the scene of internet advertisement system in high frequency, decision system, which needs to change Real-time markets, to be made
Quick response, this requires forecasting systems can provide the prediction signal about market conditions in time.
In the prior art, provide can be by the deep learning after training about the prediction signal of market conditions for forecasting system
Model directly gives, specifically: by cut-and-dried sample data training deep learning model, then using in addition a
The fitting degree of deep learning model after sample data test training, and mould is learnt by the reversed Corrected Depth of fitting result
Type so recycles, until the forecast function of deep learning model is optimal.
However, deep learning model is complex, parameter is larger, so that carrying out when prediction calculates required pair
Calculating lager time cost out is not suitable for the scenes such as high frequency transaction, internet advertisement system.In addition, passing through deep learning
Model is excessively poor come the interpretation that another major defect for providing prediction signal is exactly deep learning model, due to can not be right
The prediction signal that one model that difference can be explained provides has a good understanding, so absolute letter can not be generated to model
Appoint.
Summary of the invention
In electronic digit technical field of data processing, in order to solve the presence of Economic Forecasting Mathematical Model interpretation in the related technology
The technical problem of difference, the present invention provides a kind of construction method of Economic Forecasting Mathematical Model, device, medium and electronic equipments.
According to the one side of the application, a kind of construction method of Economic Forecasting Mathematical Model is provided, which comprises
Build Economic Forecasting Mathematical Model frame;
Initialize the parameter of each node in the Economic Forecasting Mathematical Model frame;
The training sample set for constructing Economic Forecasting Mathematical Model is obtained, includes several sample characteristics in the training sample
It is worth the desired probability distribution of prediction result corresponding with training sample;
Based on the training sample in the training sample set, the prior initialized Economic Forecasting Mathematical Model frame of training,
To obtain Economic Forecasting Mathematical Model.
According to the another aspect of the application, a kind of construction device of Economic Forecasting Mathematical Model is provided, described device includes:
Module is built, is configured as building Economic Forecasting Mathematical Model frame;
Initialization module is configured as initializing the parameter of each node in the Economic Forecasting Mathematical Model frame;
Module is obtained, is configured as obtaining the training sample set for constructing Economic Forecasting Mathematical Model;
Training module is configured as based on the training sample in the training sample set, and training is initialized in advance
Economic Forecasting Mathematical Model frame.
According to the another aspect of the application, a kind of computer-readable program medium is provided, computer program is stored with
Instruction makes computer execute foregoing method when the computer program instructions are computer-executed.
According to the another aspect of the application, a kind of electronic equipment is provided, the electronic equipment includes:
Processor;
Memory is stored with computer-readable instruction on the memory, and the computer-readable instruction is by the processing
When device executes, foregoing method is realized.
By the above technical solution of the present invention, compared with prior art, significant beneficial effect is:
(1) energy compression parameters scale, is changed into a kind of low capacity model for large capacity model, is carried out fastly with reaching to model
The speed parsing higher scene of purpose, especially requirement of real-time, such as: high frequency transaction, internet advertisement system.
(2) by the deep learning model based on relationship expression, it is extracted as a kind of top-down hierarchical model, improves mould
The interpretation of type is influenced for analyzing feature and data variation to model prediction result bring, so as to allow model solution
Release objective economy and business decision rule.
It should be understood that the above general description and the following detailed description are merely exemplary, this can not be limited
It is open.
Can be more fully appreciated from the following description in conjunction with attached drawing present invention teach that the foregoing and other aspects, reality
Apply example and feature.The feature and beneficial effect of other additional aspects such as illustrative embodiments of the invention will be retouched in following
Show in stating, or is learnt in the practice by the specific embodiment instructed according to the present invention.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows and meets implementation of the invention
Example, and be used to explain the principle of the present invention together with specification.
Fig. 1 is the block schematic illustration of Economic Forecasting Mathematical Model shown according to an exemplary embodiment;
Fig. 2 is a kind of flow chart of the construction method of Economic Forecasting Mathematical Model shown according to an exemplary embodiment;
Fig. 3 is the details flow chart of the step 240 of the embodiment shown in corresponding embodiment according to fig. 2;
Fig. 4 is the details flow chart of the step 240 of another embodiment shown in corresponding embodiment according to fig. 2;
Fig. 5 is the detail view according to the step 2405 of the embodiment shown in Fig. 3 or Fig. 4 corresponding embodiment;
Fig. 6 is a kind of block diagram of the construction device of Economic Forecasting Mathematical Model shown according to an exemplary embodiment;
Fig. 7 is a kind of electronic equipment example block diagram for realizing the above method shown according to an exemplary embodiment;
Fig. 8 is a kind of computer readable storage medium for realizing the above method shown according to an exemplary embodiment.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistented with the present invention.On the contrary, they be only with it is such as appended
The example of device and method being described in detail in claims, some aspects of the invention are consistent.
In addition, attached drawing is only the schematic illustrations of the disclosure, it is not necessarily drawn to scale.Identical attached drawing mark in figure
Note indicates same or similar part, thus will omit repetition thereof.Some block diagrams shown in the drawings are function
Energy entity, not necessarily must be corresponding with physically or logically independent entity.
The disclosure provides firstly a kind of construction method of Economic Forecasting Mathematical Model.Economic Forecasting Mathematical Model refers to and economic things
Related economic variable between (input variable and output variable) correlation theoretical construct, in the present invention, economic forecasting
Model can be for studying the functional relation to depend on each other for existence between economic phenomenon.Its purpose is to reflect the inside of economic phenomenon
Connection and its motion process, to obtain economic forecasting, aid decision making person carries out economic decision-making, and then the economy for solving reality is asked
Topic.Implementation environment of the invention can be portable mobile equipment, such as smart phone, tablet computer, laptop etc., can also
To be various fixed apparatus, for example, computer equipment, on-site terminal, desktop computer, server, work station etc..
Fig. 1 is the block schematic illustration of Economic Forecasting Mathematical Model shown according to an exemplary embodiment.
With the rapid development of science and technology, the mankind initially enter the artificial intelligence epoch, in financial transaction field, often lead to
It crosses depth learning model to handle the economic data of magnanimity, to obtain the prediction to following economic situation.But in height
In frequency transaction, carrying out economic forecasting using deep learning model is not an optimal selection, because deep learning model is more
Complexity, parameter scale is big, and the time spent by prediction process is long, can not meet the requirement in high frequency transaction to real-time, and
The interpretation of deep learning model is poor, can not make reasonable dismissal to prediction result, and therefore, it is necessary to hand over for similar to high frequency
Prediction Scenario Design one easy interpretable strong and short predicted time Economic Forecasting Mathematical Model.Accordingly, present invention employs one kind
The construction method of Economic Forecasting Mathematical Model, as shown in Figure 1, substantially it can be seen that the Economic Forecasting Mathematical Model 100 constructed by this method
It is how to execute prediction process, i.e., original economic data is input in prediction node 101, is given according to prediction node 101
Probability distribution out by probability branch 102 select a prediction child node 103, by above-mentioned initial data input selected it is pre-
It surveys in child node, a probability distribution by probability branch one prediction result 104 of selection is provided by prediction child node.
Fig. 2 is a kind of flow chart of the construction method of Economic Forecasting Mathematical Model shown according to an exemplary embodiment.Such as Fig. 2
It is shown, the method the following steps are included:
Step 210, Economic Forecasting Mathematical Model frame is built.
In one embodiment, the Economic Forecasting Mathematical Model frame can be by prediction node, probability branch and prediction result
It constitutes, wherein the prediction node is for storing a neural network model, and the probability branch is for storing by neural network model
The probability distribution for being used to select the prediction result of output.
In one embodiment, the Economic Forecasting Mathematical Model frame is also possible to predict child node, probability by prediction node
Branch and prediction result are constituted, wherein the prediction node and prediction child node are for storing neural network model, the probability branch
For storing the probability distribution for being used to select the prediction child node or prediction result by neural network model output.
Step 220, the parameter of each node in the Economic Forecasting Mathematical Model frame is initialized.
The parameter of each node is actually to be stored in each section as described above in the Economic Forecasting Mathematical Model frame
The parameter of neural network model at point, for example, prediction node 101 as shown in Figure 1 and prediction child node 103.
Initialize neural network model parameter at each node in the Economic Forecasting Mathematical Model frame mode can have it is more
Kind:
In one embodiment, neural network model parameter at each node is initialized in the Economic Forecasting Mathematical Model frame
Mode can be through zeros initialization (complete zero initialization) come what is completed.
In one embodiment, neural network model parameter at each node is initialized in the Economic Forecasting Mathematical Model frame
Mode can be by random initialization (random initializtion) come what is completed.
In one embodiment, neural network model parameter at each node is initialized in the Economic Forecasting Mathematical Model frame
Mode can be by Pre-train initialization (transfer learning initialization) come what is completed.
In one embodiment, neural network model parameter at each node is initialized in the Economic Forecasting Mathematical Model frame
Mode can be through data sensitive initialization and to complete.
In one embodiment, neural network model parameter at each node is initialized in the Economic Forecasting Mathematical Model frame
Mode can be through Xavier initialization and to complete.
In one embodiment, neural network model parameter at each node is initialized in the Economic Forecasting Mathematical Model frame
Mode can be through he initialization and to complete.
As discussed above, it should be understood that initializing in the Economic Forecasting Mathematical Model frame neural network at each node
Those of the mode of model parameter can be arbitrary, however it is not limited to go out as shown above.
Step 230, the training sample set for constructing Economic Forecasting Mathematical Model is obtained, includes several in the training sample
The desired probability distribution of sample characteristics and the corresponding prediction result of training sample.
Include several training samples in the training sample set, contains several sample characteristics in every portion training sample
Value corresponds to the desired probability distribution of prediction result with the training sample.
The main source of training sample aggregates content is that the prediction object of record has generated in transaction in history
Data, specifically, for example, building one about the purpose of A-share ticket Economic Forecasting Mathematical Model be to predict A-share ticket following 10 minutes it
Within ups and downs situation, then, for training a training sample of this Economic Forecasting Mathematical Model that can be based on following process
Obtain: firstly, statistics A-share ticket is in the status data of a certain particular state in history, this state can pass through institute as above
Several sample characteristics in the training sample mentioned are stated to embody, for example, the A-share ticket is locating a certain in history
Particular state is " stock price X1=10 yuan/strand, turnover rate X2=0.8%, transaction value X3=2.5 hundred million, capital stock in circulation X4=
12.4 hundred million, p/e ratio X5=18.8, amplitude X6=2.8%, net assets per share X7=21.6 ... ".It is specific that A-share ticket is in this
State occurred 100 times altogether in history, drop situation had occurred within 10 minutes futures wherein having 20 times, 80 times in future 10
The case where rise has occurred in minute.So according to data as above, so that it may obtain a training sample, it may be assumed that
The corresponding several sample characteristics of training sample are as follows: " stock price X1=10 yuan/strand, turnover rate X2=0.8%, at
Friendship volume X3=2.5 hundred million, capital stock in circulation X4=12.4 hundred million, p/e ratio X5=18.8, amplitude X6=2.8%, net assets per share X7=
21.6、……”
Training sample corresponds to the desired probability distribution of prediction result are as follows: " rise probability=0.8, drop probability=0.2 ".
For construct Economic Forecasting Mathematical Model training sample set acquisition modes can there are many:
In one embodiment, obtaining can be for constructing the training sample set of Economic Forecasting Mathematical Model by acquiring equipment
Or acquisition unit acquisition data and via converter convert and obtain.
In one embodiment, it obtains and is also possible to for constructing the training sample set of Economic Forecasting Mathematical Model through data
Transmission medium come read or receive staff transmission data, the transmission medium of data can be wired medium (such as count
According to line) or wireless medium (such as cordless communication network).
In one embodiment, it obtains and is also possible to climb by operation for constructing the training sample set of Economic Forecasting Mathematical Model
Worm program crawls the training sample set from transaction data base.
In one embodiment, physical hard disk can also be passed through for constructing the training sample set of Economic Forecasting Mathematical Model by obtaining
Copy obtains the training sample set, or is manually entered to obtain by staff.
It can for constructing the mode of the training sample set of Economic Forecasting Mathematical Model as discussed above, it should be understood that obtaining
Those of to be arbitrary, however it is not limited to go out as shown above.
Step 240, based on the training sample in the training sample set, the prior initialized economic forecasting mould of training
Type frame, to obtain Economic Forecasting Mathematical Model.
In one embodiment, based on the training sample in the training sample set, the prior initialized warp of training
Ji prediction model frame can be completed in the following way:
Fig. 3 is the details flow chart of the step 240 of the embodiment shown in corresponding embodiment according to fig. 2.Specifically include with
Lower step:
Step 2401, several sample characteristics in the training sample are inputted pre- in the Economic Forecasting Mathematical Model frame
The neural network model at node is surveyed, the reality of the corresponding prediction result of the training sample is exported by the neural network model
Probability distribution.
The mind at node will be predicted in several sample characteristics input Economic Forecasting Mathematical Model frame in the training sample
Through network model, by the actual probability distribution of the corresponding prediction result of neural network model output training sample.Specifically,
For example, by as described above about " the stock price X of A-share ticket1=10 yuan/strand, turnover rate X2=0.8%, transaction value X3=2.5
Hundred million, capital stock in circulation X4=12.4 hundred million, p/e ratio X5=18.8, amplitude X6=2.8%, net assets per share X7=21.6 ... "
The corresponding several sample characteristics of training sample are input to the neural network model predicted at node in Economic Forecasting Mathematical Model frame
In, the neural network model at the prediction node can then export the actual probabilities point of the corresponding prediction result of training sample
Cloth, such as: " rise probability=0.6, drop probability=0.4 ".
In one embodiment in the specific implementation, the neural network model can be BP neural network model.
Step 2402, the expected probability point of actual probability distribution prediction result corresponding with the training sample is calculated
The error amount of cloth.
In one embodiment, the expected probability of actual probability distribution prediction result corresponding with the training sample
The error amount of distribution can be the difference square of the actual probability distribution and desired probability distribution.
Specifically, for example, rise actual probabilities of the A-share ticket within following 10 minutes are 0.6, rise expected probability
It is 0.8, then the error amount is are as follows:
Error amount=(0.6-0.8)2=0.04
Step 2405, it is based on the error amount, obtains and is predicted at node for optimizing in the Economic Forecasting Mathematical Model frame
The loss function of neural network model.
In one embodiment in the specific implementation, being based on the error amount, obtain for optimizing the economic forecasting mould
It can also be performed the following steps before the loss function of neural network model at prediction node in type frame:
It is according to the detail view of the step 2405 of the embodiment shown in Fig. 3 corresponding embodiment, specifically, packet such as Fig. 5
It includes:
In the mistake for the desired probability distribution for calculating actual probability distribution prediction result corresponding with the training sample
After difference, also it should judge whether the error amount restrains, convergent standard can be preset according to actual needs, for example,
It is considered as convergence when error amount is less than 0.0001.
If it is determined that when error amount is restrained, then deconditioning;
If it is determined that error does not restrain, then obtains and predict nerve at node in the Economic Forecasting Mathematical Model frame for optimizing
The loss function of network model continues to train.
It is noted herein that under normal circumstances, the Economic Forecasting Mathematical Model frame is needed by several training samples
It is multiple training can be only achieved the convergent requirement of error amount.
In one embodiment in the specific implementation, the loss function of the acquisition may is that after minimum processing
Wherein, pmIndicating m has inclined expert model probability distribution, pgIndicate the probability distribution of the extensive model of broad sense, q is indicated
The destination probability of extraction is distributed.
Step 2406, it is based on the loss function, reversely updates and predicts mind at node in the Economic Forecasting Mathematical Model frame
Parameter through network model, with the neural network model at Optimization Prediction node.
In another embodiment, based on the training sample in the training sample set, training is initialized in advance
Economic Forecasting Mathematical Model frame can be completed in the following way:
Fig. 4 is the details flow chart of the step 240 of another embodiment shown in corresponding embodiment according to fig. 2.It specifically includes
Following steps:
Step 2403, several sample characteristics in the training sample are inputted pre- in the Economic Forecasting Mathematical Model frame
The neural network model at node and at prediction child node is surveyed, by the neural network model at prediction node and prediction child node
The actual probability distribution of the corresponding prediction child node of the training sample and the actual probability distribution of prediction result are exported respectively.
Several sample characteristics in the training sample are inputted in the Economic Forecasting Mathematical Model frame and are predicted at node
With the neural network model at prediction child node, exported respectively by the neural network model at prediction node and prediction child node
The actual probability distribution of the corresponding prediction child node of training sample and the actual probability distribution of prediction result.Specifically, for example, will
As described above about " the stock price X of A-share ticket1=10 yuan/strand, turnover rate X2=0.8%, transaction value X3=2.5 hundred million, it circulates
Capital stock X4=12.4 hundred million, p/e ratio X5=18.8, amplitude X6=2.8%, net assets per share X7=21.6, training sample ... "
Corresponding several sample characteristics, which are input in Economic Forecasting Mathematical Model frame as shown in Figure 1, to be predicted at node and prediction child node
Neural network model in, it is described prediction node at and prediction child node at neural network model then respectively export training sample
The actual probability distribution of this corresponding prediction child node and the actual probability distribution of prediction result.Such as:
Wherein, the actual probability distribution of the corresponding prediction child node of output training sample are as follows: " rise probability (probability branch 2)
=0.6, drop probability (probability branch 1)=0.4 ";
Predict the actual probability distribution of the corresponding prediction result of neural network model output training sample at child node 1
Are as follows: " 1 probability of prediction result (probability branch 3)=0.3,2 probability of prediction result (probability branch 4)=0.7 ";
Predict the actual probability distribution of the corresponding prediction result of neural network model output training sample at child node 2
Are as follows: " 3 probability of prediction result (probability branch 5)=0.8,4 probability of prediction result (probability branch 6)=0.2 ".
In one embodiment in the specific implementation, the neural network model can be BP neural network model.
Step 2404, it is general to calculate separately the reality that at the prediction node and neural network model exports at prediction child node
The error amount of rate distribution prediction child node and the corresponding desired probability distribution of prediction result corresponding with the training sample.
It is noted herein that in the present embodiment, also including in training sample described in step 230 as described above
The desired probability distribution of the corresponding prediction child node of training sample.
For example, such as Fig. 1, the desired probability distribution of the corresponding prediction child node of training sample are as follows: " rise probability (probability branch 2)
=0.8, drop probability (probability branch 1)=0.2 ";
Training sample corresponds to the desired probability distribution of prediction result 1 are as follows: and " 1 probability of prediction result (probability branch 3)=0.4 ",
Training sample corresponds to the desired probability distribution of prediction result 2 are as follows: " 2 probability of prediction result (probability branch 4)=0.6 ";
Training sample corresponds to the desired probability distribution of prediction result 3 are as follows: and " 3 probability of prediction result (probability branch 5)=0.7 ",
Training sample corresponds to the desired probability distribution of prediction result 4 are as follows: " 4 probability of prediction result (probability branch 6)=0.3 ".
In one embodiment, neural network model output at the prediction node calculated separately and at prediction child node
Actual probability distribution it is corresponding with training sample prediction child node and the corresponding desired probability distribution of prediction result error amount
It can be the difference square of the actual probability distribution and desired probability distribution.
Specifically, for example, rise (probability branch 2) actual probabilities of the A-share ticket within following 10 minutes are 0.6, on
The expected probability that rises is (probability branch 2) 0.8, then the error amount is are as follows:
Error amount=(0.6-0.8)2=0.04
Further for example, at prediction child node 1, the error amount are as follows:
Error amount=(0.7-0.6)2=(0.3-0.4)2=0.01
Further for example, at prediction child node 2, the error amount are as follows:
Error amount=(0.8-0.7)2=(0.2-0.3)2=0.01
Step 2405, be based on the error amount, obtain for optimize in the Economic Forecasting Mathematical Model frame predict node and
Predict the loss function of neural network model at child node.
In one embodiment in the specific implementation, being based on the error amount, obtain for optimizing Economic Forecasting Mathematical Model frame
Predict that node with before the loss function of neural network model at prediction child node, can also perform the following steps in frame:
It is according to the detail view of the step 2405 of the embodiment shown in Fig. 3 corresponding embodiment, specifically, packet such as Fig. 5
It includes:
In the mistake for the desired probability distribution for calculating actual probability distribution prediction result corresponding with the training sample
After difference, also it should judge whether the error amount restrains, convergent standard can be preset according to actual needs, example
Such as, error amount less than 0.0001 when be considered as convergence.
If it is determined that when error amount is restrained, then deconditioning;
If it is determined that error does not restrain, then obtains and predict node and prediction in the Economic Forecasting Mathematical Model frame for optimizing
The loss function of neural network model at child node, continues to train.
It is noted herein that under normal circumstances, the Economic Forecasting Mathematical Model frame is needed by several training samples
It is multiple training can be only achieved the convergent requirement of error amount.
In one embodiment in the specific implementation, the loss function of the acquisition may is that after minimum processing
Wherein, pmIndicating m has inclined expert model probability distribution, pgIndicate the probability distribution of the extensive model of broad sense, q is indicated
The destination probability of extraction is distributed.
Step 2406, it is based on the loss function, reversely updates and predicts node and pre- in the Economic Forecasting Mathematical Model frame
The parameter of neural network model at child node is surveyed, with the neural network model at Optimization Prediction node and prediction child node.
In conclusion capableing of the parameter scale of compact model by the method, Conventional mass model is changed into one kind
Low capacity model carries out the higher scene of fast resolving purpose, especially requirement of real-time to model to reach.The present invention will close
Deep learning model based on system's expression, is extracted as a kind of top-down hierarchical model, improves the interpretation of model, uses
It is influenced in analysis feature and data variation to model prediction result bring, so as to make model explanation objective economical and business
Decision laws.
It is the device of the invention embodiment below.
The disclosure additionally provides a kind of construction device of Economic Forecasting Mathematical Model.Fig. 6 is shown according to an exemplary embodiment
A kind of Economic Forecasting Mathematical Model construction device block diagram.As shown in fig. 6, device 600 includes:
Module 610 is built, is configured as building Economic Forecasting Mathematical Model frame;
Initialization module 620 is configured as initializing the parameter of each node in the Economic Forecasting Mathematical Model frame;
Module 630 is obtained, is configured as obtaining the training sample set for constructing Economic Forecasting Mathematical Model;
Training module 640 is configured as based on the training sample in the training sample set, and training has initialized in advance
Economic Forecasting Mathematical Model frame.
According to the third aspect of the disclosure, a kind of electronic equipment that can be realized the above method is additionally provided.
Person of ordinary skill in the field it is understood that various aspects of the invention can be implemented as system, method or
Program product.Therefore, various aspects of the invention can be embodied in the following forms, it may be assumed that complete hardware embodiment, complete
The embodiment combined in terms of full Software Implementation (including firmware, microcode etc.) or hardware and software, can unite here
Referred to as circuit, " module " or " system ".
The electronic equipment 1000 of this embodiment according to the present invention is described referring to Fig. 7.The electronics that Fig. 7 is shown
Equipment 1000 is only an example, should not function to the embodiment of the present invention and use scope bring any restrictions.
As shown in fig. 7, electronic equipment 1000 is showed in the form of universal computing device.The component of electronic equipment 1000 can be with
Including but not limited to: at least one above-mentioned processing unit 1010, connects not homologous ray group at least one above-mentioned storage unit 1020
The bus 1030 of part (including storage unit 1020 and processing unit 1010).
Wherein, the storage unit is stored with program code, and said program code can be held by the processing unit 1010
Row, so that various according to the present invention described in the execution of the processing unit 1010 above-mentioned " embodiment method " part of this specification
The step of illustrative embodiments.
Storage unit 1020 may include the readable medium of volatile memory cell form, such as Random Access Storage Unit
(RAM) 1021 and/or cache memory unit 1022, it can further include read-only memory unit (ROM) 1023.
Storage unit 1020 can also include program/utility with one group of (at least one) program module 1025
1024, such program module 1025 includes but is not limited to: operating system, one or more application program, other program moulds
It may include the realization of network environment in block and program data, each of these examples or certain combination.
Bus 1030 can be to indicate one of a few class bus structures or a variety of, including storage unit bus or storage
Cell controller, peripheral bus, graphics acceleration port, processing unit use any bus structures in a variety of bus structures
Local bus.
Electronic equipment 1000 can also be with one or more external equipments 1200 (such as keyboard, sensing equipment, bluetooth equipment
Deng) communication, can also be enabled a user to one or more equipment interact with the electronic equipment 1000 communicate, and/or with make
The electronic equipment 1000 can with it is one or more of the other calculating equipment be communicated any equipment (such as router, modulation
Demodulator etc.) communication.This communication can be carried out by input/output (I/O) interface 1050.Also, electronic equipment 1000
Network adapter 1060 and one or more network (such as local area network (LAN), wide area network (WAN) and/or public affairs can also be passed through
Common network network, such as internet) communication.As shown, network adapter 1060 passes through its of bus 1030 and electronic equipment 1000
The communication of its module.It should be understood that although not shown in the drawings, other hardware and/or software can be used in conjunction with electronic equipment 1000
Module, including but not limited to: microcode, device driver, redundant processing unit, external disk drive array, RAID system, magnetic
Tape drive and data backup storage system etc..
Through the above description of the embodiments, those skilled in the art is it can be readily appreciated that example described herein is implemented
Mode can also be realized by software realization in such a way that software is in conjunction with necessary hardware.Therefore, according to the disclosure
The technical solution of embodiment can be embodied in the form of software products, which can store non-volatile at one
Property storage medium (can be CD-ROM, USB flash disk, mobile hard disk etc.) in or network on, including some instructions are so that a calculating
Equipment (can be personal computer, server, terminal installation or network equipment etc.) is executed according to disclosure embodiment
Method.
According to the fourth aspect of the disclosure, a kind of computer readable storage medium is additionally provided, being stored thereon with can be real
The program product of existing this specification above method.In some possible embodiments, various aspects of the invention can also be real
It is now a kind of form of program product comprising program code, when described program product is run on the terminal device, the journey
Sequence code is each according to the present invention described in above-mentioned " illustrative methods " part of this specification for executing the terminal device
The step of kind illustrative embodiments.
Refering to what is shown in Fig. 8, describing the program product for realizing the above method of embodiment according to the present invention
1100, can using portable compact disc read only memory (CD-ROM) and including program code, and can in terminal device,
Such as it is run on PC.However, program product of the invention is without being limited thereto, in this document, readable storage medium storing program for executing can be with
To be any include or the tangible medium of storage program, the program can be commanded execution system, device or device use or
It is in connection.
Described program product can be using any combination of one or more readable mediums.Readable medium can be readable letter
Number medium or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can be but be not limited to electricity, magnetic, optical, electromagnetic, infrared ray or
System, device or the device of semiconductor, or any above combination.The more specific example of readable storage medium storing program for executing is (non exhaustive
List) include: electrical connection with one or more conducting wires, portable disc, hard disk, random access memory (RAM), read-only
Memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read only memory
(CD-ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
In carry readable program code.The data-signal of this propagation can take various forms, including but not limited to electromagnetic signal,
Optical signal or above-mentioned any appropriate combination.Readable signal medium can also be any readable Jie other than readable storage medium storing program for executing
Matter, the readable medium can send, propagate or transmit for by instruction execution system, device or device use or and its
The program of combined use.
The program code for including on readable medium can transmit with any suitable medium, including but not limited to wirelessly, have
Line, optical cable, RF etc. or above-mentioned any appropriate combination.
The program for executing operation of the present invention can be write with any combination of one or more programming languages
Code, described program design language include object oriented program language-Java, C++ etc., further include conventional
Procedural programming language-such as " C " language or similar programming language.Program code can be fully in user
It calculates and executes in equipment, partly executes on a user device, being executed as an independent software package, partially in user's calculating
Upper side point is executed on a remote computing or is executed in remote computing device or server completely.It is being related to far
Journey calculates in the situation of equipment, and remote computing device can pass through the network of any kind, including local area network (LAN) or wide area network
(WAN), it is connected to user calculating equipment, or, it may be connected to external computing device (such as utilize ISP
To be connected by internet).
In addition, above-mentioned attached drawing is only the schematic theory of processing included by method according to an exemplary embodiment of the present invention
It is bright, rather than limit purpose.It can be readily appreciated that the time that above-mentioned processing shown in the drawings did not indicated or limited these processing is suitable
Sequence.In addition, be also easy to understand, these processing, which can be, for example either synchronously or asynchronously to be executed in multiple modules.
It should be understood that the present invention is not limited to the precise structure already described above and shown in the accompanying drawings, and
And various modifications and change can executed without departing from the scope.The scope of the present invention is limited only by the attached claims.
Claims (10)
1. a kind of construction method of Economic Forecasting Mathematical Model characterized by comprising
Build Economic Forecasting Mathematical Model frame;
Initialize the parameter of each node in the Economic Forecasting Mathematical Model frame;
Obtain the training sample set for constructing Economic Forecasting Mathematical Model, in the training sample comprising several sample characteristics and
The desired probability distribution of the corresponding prediction result of training sample;
Based on the training sample in the training sample set, the prior initialized Economic Forecasting Mathematical Model frame of training, with
To Economic Forecasting Mathematical Model.
2. the method according to claim 1, wherein the Economic Forecasting Mathematical Model frame is by prediction node, probability
Branch and prediction result constitute, wherein the prediction node for store a neural network model, the probability branch for storage by
The probability distribution for being used to select the prediction result of the neural network model output.
3. according to the method described in claim 2, it is characterized in that, the training sample based in the training sample set
This, the prior initialized Economic Forecasting Mathematical Model frame of training, comprising:
Several sample characteristics in the training sample are inputted in the Economic Forecasting Mathematical Model frame to the mind predicted at node
Through network model, the actual probability distribution of the corresponding prediction result of the training sample is exported by the neural network model;
Calculate the error amount of the desired probability distribution of actual probability distribution prediction result corresponding with the training sample;
Based on the error amount, obtains and predict neural network model at node in the Economic Forecasting Mathematical Model frame for optimizing
Loss function;
Based on the loss function, the ginseng that neural network model at node is predicted in the Economic Forecasting Mathematical Model frame is reversely updated
Number, with the neural network model at Optimization Prediction node.
4. the method according to claim 1, wherein the Economic Forecasting Mathematical Model frame is by prediction node, prediction
Child node, probability branch and prediction result are constituted, wherein the prediction node and prediction child node are for storing neural network mould
Type, the probability branch, which is used to store, described predicts the general of child node or prediction result for selecting by neural network model output
Rate distribution.
5. according to the method described in claim 4, it is characterized in that, also corresponding pre- comprising training sample in the training sample
The desired probability distribution of child node is surveyed,
The training sample based in the training sample set, the prior initialized Economic Forecasting Mathematical Model frame of training,
Include:
Several sample characteristics in the training sample are inputted at prediction node and pre- in the Economic Forecasting Mathematical Model frame
Survey child node at neural network model, by prediction node and prediction child node at neural network model export respectively described in
The actual probability distribution of the corresponding prediction child node of training sample and the actual probability distribution of prediction result;
Calculate separately it is described prediction node at and prediction child node at neural network model export actual probability distribution with it is described
The error amount of the corresponding prediction child node and the corresponding desired probability distribution of prediction result of training sample;
Based on the error amount, obtains and predict mind at node and prediction child node in the Economic Forecasting Mathematical Model frame for optimizing
Loss function through network model;
Based on the loss function, reversely updates and predict nerve at node and prediction child node in the Economic Forecasting Mathematical Model frame
The parameter of network model, with the neural network model at Optimization Prediction node and prediction child node.
6. the method according to claim 3 or 5, which is characterized in that it is described to be based on the error amount, it obtains for optimizing warp
Node is predicted in Ji prediction model frame or predicts the loss function of neural network model at child node, comprising:
Judge whether the error amount restrains;
If the error amount is not restrained, obtains and predict node or prediction child node in Economic Forecasting Mathematical Model frame for optimizing
Locate the loss function of neural network model, to continue to train;
If the error amount convergence, deconditioning.
7. the method according to claim 3 or 5, which is characterized in that the neural network model is BP neural network model.
8. a kind of construction device of Economic Forecasting Mathematical Model, which is characterized in that described device includes:
Module is built, for building Economic Forecasting Mathematical Model frame;
Initialization module, for initializing the parameter of each node in the Economic Forecasting Mathematical Model frame;
Module is obtained, for obtaining the training sample set for constructing Economic Forecasting Mathematical Model;
Training module, for based on the training sample in the training sample set, training initialized economic forecasting in advance
Model framework.
9. a kind of computer-readable program medium, which is characterized in that it is stored with computer program instructions, when the computer journey
When sequence instruction is computer-executed, computer is made to execute method according to any one of claim 1 to 7.
10. a kind of building electronic equipment of Economic Forecasting Mathematical Model, which is characterized in that the electronic equipment includes:
Processor;
Memory is stored with computer-readable instruction on the memory, and the computer-readable instruction is held by the processor
When row, method as described in any one of claim 1 to 7 is realized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910273964.6A CN110110894A (en) | 2019-04-04 | 2019-04-04 | Construction method, device, medium, the electronic equipment of Economic Forecasting Mathematical Model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910273964.6A CN110110894A (en) | 2019-04-04 | 2019-04-04 | Construction method, device, medium, the electronic equipment of Economic Forecasting Mathematical Model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110110894A true CN110110894A (en) | 2019-08-09 |
Family
ID=67485217
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910273964.6A Pending CN110110894A (en) | 2019-04-04 | 2019-04-04 | Construction method, device, medium, the electronic equipment of Economic Forecasting Mathematical Model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110110894A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111709532A (en) * | 2020-05-26 | 2020-09-25 | 重庆大学 | Model-independent local interpretation-based online shopping representative sample selection system |
-
2019
- 2019-04-04 CN CN201910273964.6A patent/CN110110894A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111709532A (en) * | 2020-05-26 | 2020-09-25 | 重庆大学 | Model-independent local interpretation-based online shopping representative sample selection system |
CN111709532B (en) * | 2020-05-26 | 2023-09-22 | 重庆大学 | Online shopping representative sample selection system based on model-independent local interpretation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109657805A (en) | Hyper parameter determines method, apparatus, electronic equipment and computer-readable medium | |
CN106095942B (en) | Strong variable extracting method and device | |
Assefa Tsehayae et al. | Developing and optimizing context-specific fuzzy inference system-based construction labor productivity models | |
CN107169534A (en) | Model training method and device, storage medium, electronic equipment | |
CN109242033A (en) | Wafer defect method for classifying modes and device, storage medium, electronic equipment | |
Zhao et al. | Revolutionizing finance with llms: An overview of applications and insights | |
CN110111139A (en) | Behavior prediction model generation method, device, electronic equipment and readable medium | |
CN110009480A (en) | The recommended method in judicial collection path, device, medium, electronic equipment | |
Pupara et al. | An institution recommender system based on student context and educational institution in a mobile environment | |
CN114511152A (en) | Training method and device of prediction model | |
CN110704803A (en) | Target object evaluation value calculation method and device, storage medium and electronic device | |
Yang et al. | Predicting student learning effectiveness in higher education based on big data analysis | |
CN110033165A (en) | The recommended method of overdue loaning bill collection mode, device, medium, electronic equipment | |
CN114331380A (en) | Method, system, equipment and storage medium for predicting occupational flow relationship | |
Yuan et al. | EFFC-Net: lightweight fully convolutional neural networks in remote sensing disaster images | |
CN110110894A (en) | Construction method, device, medium, the electronic equipment of Economic Forecasting Mathematical Model | |
CN117113222A (en) | Data analysis model generation method and device and electronic equipment | |
Zhu et al. | A poverty index prediction model for students based on PSO-LightGBM | |
CN109542795A (en) | Recommend method, apparatus, the medium, electronic equipment of test action for user | |
CN108984680B (en) | Information recommendation method and device, server and storage medium | |
Stefan et al. | Alternative Methods for Business Process Planning | |
Hao et al. | An enhanced two phase estimation of distribution algorithm for solving scheduling problem | |
CN111179070A (en) | Loan risk timeliness prediction system and method based on LSTM | |
Hogo | The design of academic programs using rough set association rule mining | |
CN114943274B (en) | Model training method, device, storage medium, server, terminal and system |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190809 |