CN108198072A - A kind of system of artificial intelligence assessment financial product feature - Google Patents
A kind of system of artificial intelligence assessment financial product feature Download PDFInfo
- Publication number
- CN108198072A CN108198072A CN201711482971.4A CN201711482971A CN108198072A CN 108198072 A CN108198072 A CN 108198072A CN 201711482971 A CN201711482971 A CN 201711482971A CN 108198072 A CN108198072 A CN 108198072A
- Authority
- CN
- China
- Prior art keywords
- model
- financial product
- deep learning
- data
- learning 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Technology Law (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of artificial intelligence assessment financial product feature system, including：Data acquisition module is used for：Acquire predetermined time before on financial product predetermined time the characteristic information influential data of tool and financial product predetermined time characteristic information as training sample, acquire before prediction time on financial product prediction time the characteristic information influential data of tool as test sample, price and/or behavior of the characteristic information including the financial product of financial product；Model training module is used for：Initial deep learning model is obtained, and deep learning model is trained using training sample, obtains the deep learning model with forecast function；Data prediction module, is used for：Using test sample as the input of deep learning model, predicted characteristics information of the financial product in prediction time of deep learning model output is obtained.The prediction of financial product price and/or behavior is realized in this way, improves prediction accuracy.
Description
Technical field
The present invention relates to deep learning technology fields, special more specifically to a kind of artificial intelligence assessment financial product
The system of sign.
Background technology
Financial product refers to the various carriers of financing process, it includes currency, gold, foreign exchange, marketable securities etc..It is exactly
It says, these financial products are exactly the dealing object in financial market, and supply and demand both sides form financial product valency by market competition principle
Lattice such as interest rate or earning rate, are finally completed transaction, achieve the purpose that circulate necessary funds.
If the prediction to information such as the price of a certain moment financial product or behaviors can be realized, have for financial industry
There is great meaning.It is usually the price for realizing financial product by its professional standing and experience by professional in the prior art
Or the information predictions such as behavior, but this mode, easily by the subjective impact of professional, accuracy is relatively low.
In conclusion how to provide a kind of price of the higher financial product of prediction accuracy or behavior prediction scheme, it is
Those skilled in the art's urgent problem to be solved at present.
Invention content
The object of the present invention is to provide a kind of systems of artificial intelligence assessment financial product feature, can improve financial product
Price or behavior prediction accuracy.
To achieve these goals, the present invention provides following technical solution：
A kind of system of artificial intelligence assessment financial product feature, including：
Data acquisition module is used for：Have before acquisition predetermined time to characteristic information of the financial product in predetermined time
The data of influence and financial product predetermined time characteristic information as training sample, finance is produced before acquisition prediction time
Characteristic information of the product in prediction time has influential data as test sample, and the characteristic information of the financial product includes should
The price of financial product and/or behavior；
Model training module is used for：Initial deep learning model is obtained, and using the training sample to the depth
Learning model is trained, and obtains the deep learning model with forecast function；
Data prediction module, is used for：Using the test sample as the input of the deep learning model, the depth is obtained
Spend predicted characteristics information of the financial product in prediction time of learning model output.
Preferably, it further includes：
Model chooses module, is used for：The model for obtaining extraneous input chooses instruction, and determines to choose with the model and instruct
Corresponding model structure and model parameter；
Super ginseng preset module, is used for：It determines hyper parameter corresponding with the model structure, and is selected based on predetermined manner
The value of each hyper parameter, obtains the initial deep learning model for including the model structure, model parameter and hyper parameter.
Preferably, the super ginseng setup module includes：
Super ginseng setting unit, for utilizing the value for emulating each hyper parameter of determining method selection.
Preferably, the model training module includes：
Model training unit, is used for：The framework for obtaining external world's input chooses instruction, determines corresponding with framework selection instruction
Program architecture, program code corresponding with described program framework is generated, and described in utilization based on initial deep learning model
Program code is trained initial deep learning model based on the training sample.
Preferably, the model training module further includes：
Program optimization unit, is used for：The accuracy and prediction of the deep learning model are obtained based on the training sample
Speed, judges whether the accuracy and predetermined speed reach preset requirement, if it is, determining the deep learning model
Training complete, if it is not, then adjustment said program code, and indicate that the model training unit performs and utilize described program generation
The step of code training deep learning model, until the accuracy of the deep learning model and predetermined speed reach preset requirement and be
Only.
Preferably, it further includes：
Data preprocessing module is used for：Deduplication operation is carried out to the training sample and the test sample respectively and is gone
It makes an uproar operation, and the training sample and the test sample is stored into unified data storage.
Preferably, it further includes：
Data structured module, is used for：To there are the data of identical meanings in the training sample and the test sample
Unified character is converted to, the training sample converted and test sample are structured as the data with unified form, and
Quality testing is carried out to the training sample after structuring and test sample, if detection acquired results meet expected results, really
Fixed number is according to acquiring successfully, otherwise, it indicates that the data acquisition module resurveys data.
Preferably, the data prediction module includes：
Data conversion unit, is used for：The predicted characteristics information is converted into the structure of extraneous setting.
Preferably, it further includes：
Aid decision module, is used for：The predicted characteristics information that the data prediction module is obtained is according to default output side
Formula is arranged, and is shown to arranging acquired results.
Preferably, the aid decision module further includes：
Feedback capture unit, is used for：Receive actual characteristic of the financial product in the prediction time of extraneous input
Information, and incremental learning is carried out to the deep learning model using the actual characteristic information and corresponding test sample, with profit
It is predicted with the deep learning model realization characteristic information after incremental learning is carried out.
A kind of system of artificial intelligence assessment financial product feature provided by the invention, including：Data acquisition module is used
In：Acquire predetermined time before on financial product predetermined time the characteristic information influential data of tool and financial product pre-
If the characteristic information at moment has characteristic information of the financial product in prediction time before acquisition prediction time as training sample
Influential data include price and/or the behavior of the financial product as test sample, the characteristic information of the financial product；
Model training module is used for：Initial deep learning model is obtained, and using the training sample to the deep learning model
It is trained, obtains the deep learning model with forecast function；Data prediction module, is used for：Using the test sample as
The input of the deep learning model obtains prediction of the financial product in prediction time of the deep learning model output
Characteristic information.In technical solution provided in an embodiment of the present invention, by acquiring predetermined time before to financial product when default
The characteristic information at quarter has the characteristic information of influential data and financial product in predetermined time as training sample, and utilizes sample
This information trains to obtain deep learning model, and then produce finance using before prediction time of the deep learning model based on acquisition
Characteristic information of the product in prediction time has influential data, and characteristic information of the financial product in prediction time is predicted,
Corresponding prediction characteristic information is obtained, and characteristic information includes behavior and/or the price of product；So as to pass through this artificial intelligence
Mode realize that financial product in certain moment price and/or the prediction of behavior, is avoided and realized in the prior art by professional
Since the relatively low situation of accuracy caused by subjective impact occurs during above-mentioned prediction, financial product feature is improved so as to reach
Prediction accuracy.
Description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention, for those of ordinary skill in the art, without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is the first structure for the system that a kind of artificial intelligence provided in an embodiment of the present invention assesses financial product feature
Schematic diagram；
Fig. 2 is second of structure of the system that a kind of artificial intelligence provided in an embodiment of the present invention assesses financial product feature
Schematic diagram.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other without making creative work
Embodiment shall fall within the protection scope of the present invention.
Referring to Fig. 1, it is it illustrates a kind of artificial intelligence assessment financial product feature provided in an embodiment of the present invention
The structure diagram of system can include：
Data acquisition module 11, is used for：Have before acquiring predetermined time to characteristic information of the financial product in predetermined time
Influential data and financial product predetermined time characteristic information as training sample, to finance before acquisition prediction time
Characteristic information of the product in prediction time has influential data as test sample, and the characteristic information of financial product includes the gold
Melt price and/or the behavior of product；
Model training module 12, is used for：Initial deep learning model is obtained, and using training sample to deep learning mould
Type is trained, and obtains the deep learning model with forecast function；
Data prediction module 13, is used for：Using test sample as the input of deep learning model, deep learning model is obtained
The financial product of output is in the predicted characteristics information of prediction time.
Training sample is the data having determined in history, and predetermined time is a certain moment in history, and prediction time is
At the time of needing to predict the characteristic information of financial product, predetermined time and prediction time can according to actual needs into
Row determines.It is pre- in the characteristic information influential data of tool of predetermined time and acquisition on financial product before acquiring predetermined time
Can be specifically that acquisition is corresponding pre- on characteristic information tool influential data of the financial product in prediction time before surveying the moment
If the data in the preset time period set according to actual needs before moment or prediction time.So as to be realized by training sample
The training of deep learning model is believed with the characteristic of deep learning model prediction prediction time financial product obtained using training
Breath.Wherein, the price of financial product is the value dimension of financial product, and the behavior of financial product can be that financial product can
The state that can occur or operation etc., if financial product is stock, the price of financial product is then stock price, the row of financial product
To be then stock price tendency, that is, the price of financial product and behavior and the realization for corresponding to technical solution in the prior art are former
Reason is consistent, and details are not described herein.
It should be noted that financial product (such as the industry exchange rate, the trend of stock prices) is dependent on existing society, politics, economy
And situations such as enterprise operation, therefore when can setting existing society, politics, economic and enterprise operation according to actual needs
Certain specific factors be on the influential influence factor of financial product tool (such as enterprise operation, war event), and to finance production
The characteristic information of product at a time has the data that influential data can be then preset influence factor.It is specific next
It says, is acquired in the application and the influential data of financial product tool can be specifically acquired by following several ways：Network,
The approach such as TV, newspaper acquisition magnanimity can effectively influence the data of financial product；Annual report, financial report of each listed company etc. are open to pass through
Seek the data of each company management situation of information collection；The current and/or price of moment financial product before.It certainly can also basis
Actual needs carries out other settings, within protection scope of the present invention.
In addition, assessment system disclosed in the present application is specifically as follows assessment system end to end, and specifically, the two of system
End is respectively the one end for realizing data acquisition and the one end for realizing forecast function, and in general, data acquisition module is set to reality
One end of existing data acquisition, and other modules are set to the one end for realizing forecast function, thus by artificial intelligence end to end
Assessment system realizes the feature evaluation to financial product.
In technical solution provided in an embodiment of the present invention, by acquiring predetermined time before to financial product in predetermined time
Characteristic information have the characteristic information of influential data and financial product in predetermined time as training sample, and utilize sample
Information trains to obtain deep learning model, and then using prediction time of the deep learning model based on acquisition before to financial product
Have influential data in the characteristic information of prediction time, characteristic information of the financial product in prediction time is predicted, is obtained
To corresponding prediction characteristic information, and characteristic information includes behavior and/or the price of product；So as to pass through this artificial intelligence
Mode realizes that financial product in certain moment price and/or the prediction of behavior, is avoided and realized in the prior art by professional
Since the relatively low situation of accuracy caused by subjective impact occurs when stating prediction, the pre- of financial product feature is improved so as to reach
Survey accuracy.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, can also include：
Model chooses module, is used for：The model for obtaining external world's input chooses instruction, and determines corresponding with model selection instruction
Model structure and model parameter；
Super ginseng preset module, is used for：It determines hyper parameter corresponding with model structure, and is selected based on predetermined manner each
The value of hyper parameter obtains the initial deep learning model for including model structure, model parameter and hyper parameter.
It sets and is input to according to actual needs it should be noted that initial deep learning model can be staff
It is in system or needed by system based on user and application scenarios automatically generate, so that deep learning model
More meet user demand, be more suitable for current application scene, further improve the prediction accuracy of financial product feature.Tool
For body, system can provide different modularization deep learning model structures (abbreviation model structure) and be chosen for user,
And system offer model structure can be while the mode of icon shows different model structures, and pay close attention to a certain model in user
It is provided during structure and the information that the model structure is specifically introduced is used as with reference to information；Wherein user pays close attention to a certain model structure
Can be that mouse is moved to the model structure or is realized by other modes by user；And have to model structure
The information of body introduction can include simplification figure, parameter introduction, function introduction of model structure etc., specifically can be according to actual needs
It is set, details are not described herein.After user chooses multiple model structures, model structure that system is chosen automatically according to user
According to the sequencing automatic butt that user chooses to different practical model structures, it is made of finally these model structures
In model structure, i.e. the application the corresponding model structure of instruction is chosen with model.In deep learning algorithm, to different models
Structure needs to define different parameters, i.e. model parameter, and it is mould at the interface that user interface can be provided in the application for user
The parameters of type structure are defined, so as to obtain choosing the corresponding model parameter of instruction with model.Wherein, model selection refers to
Enable the operation that can include that user chooses any and model structure that system is realized and model parameter is chosen.Furthermore it is also possible to
Model parameter is chosen automatically based on the technologies such as sampling are uniformly distributed, so as to extrapolate optimal model parameter automatically, and
The model parameter extrapolated is shown and is referred to for user in Definition Model parameter, so as to further ensure the reliable of parameter
Property.
In addition, for the model structure and model parameter that have had determined, it is also necessary to which entire deep learning model is carried out
Hyper parameter defines, it is necessary first to determine need which hyper parameter set for the model structure having had determined, then utilize warp
The value that the modes such as selection or emulation judgement select each hyper parameter is tested, wherein for by emulating the super ginseng for judging to determine
Several values can determine optimal hyper parameter value according to model parameter and model structure, certainly can also be according to actual needs
The selection of hyper parameter is realized using other modes, within protection scope of the present invention.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, surpassing ginseng setup module can be with
Including：
Super ginseng setting unit, for utilizing the value for emulating each hyper parameter of determining method selection.
It is preferably determined in the application using the value for emulating determining method and determining each hyper parameter by multiple emulation
Go out the best results that each hyper parameter causes deep learning model in what specific value, so as to further ensure deep learning
The forecasting accuracy of model.It is further to note that the definition of model structure, model parameter and hyper parameter is with showing in the application
Have that the definition that title is corresponded in technology is identical, and details are not described herein.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, model training module can be with
Including：
Model training unit, is used for：The framework for obtaining external world's input chooses instruction, determines corresponding with framework selection instruction
Program architecture, program code corresponding with program architecture is generated, and utilize program code based on initial deep learning model
Initial deep learning model is trained based on training sample.
System can provide multiple and different program architecture (i.e. software architecture), sequencing to be facilitated to realize deep learning mould
The training operation of type, specifically, the program architecture that can be chosen to user's displaying in a manner of list in the application, from
And when user selectes a certain program architecture, generate program generation corresponding with initial depth network model according to the program architecture
Code, and then program code execution realizes the training of deep learning model.Wherein, program code can call training in the process of running
Sample realizes corresponding training function to be based on training sample；And for the selection of program architecture, it can also be sharp according to actual needs
It realizes in other ways, as the selection based on previous user determines the most program architecture of access times as finally choosing
Program architecture selects what the program architecture chosen under identical application scenarios before current time was the most finally chosen
Program architecture etc., within protection scope of the present invention.So as to realize the training of deep learning model using program in machine code, make
It obtains the model and reaches preferably state.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, model training module may be used also
To include：
Program optimization unit, is used for：Accuracy and predetermined speed of deep learning model are obtained based on training sample, is judged
Whether accuracy and predetermined speed reach preset requirement, if it is, determine that the training of deep learning model is completed, if
It is no, then adjustment programme code, and indicate the step of model training unit is performed using program code training deep learning model, directly
Until the accuracy of deep learning model and predetermined speed reach preset requirement.
It should be noted that can using in one group of training sample on the price of the financial product influential data of tool as deep
Spend the input of learning model, the characteristic information exported, by the data and the financial product in that group of training sample of input
Characteristic information be compared, the similarity of two characteristic informations, as accuracy can be obtained, if the accuracy reaches pre-
The value (i.e. the value of accuracy in preset requirement) of the accuracy first set then illustrates that the accuracy of deep learning model reaches and wants
It asks；Meanwhile by using in one group of training sample on the influential data of the price of financial product tool as the defeated of deep learning model
Start timing at the time of entering to the characteristic information moment stopping timing being exported, the inverse for determining duration obtained by timing is prediction
Speed if the predetermined speed reaches the value (i.e. the value of predetermined speed in preset requirement) of preset predetermined speed, is said
Predetermined speed of bright deep learning model reaches requirement.When predetermined speed and accuracy reach requirement, depth can be utilized
Learning model realizes forecast function, and otherwise, then model parameter and/or the value of hyper parameter in adjustment programme code, deep up to causing
Until the speed of degree learning model and accuracy reach requirement, so as to ensure that deep learning model has preferably performance.
Wherein, realizing that accuracy and when acquisition of predetermined speed can also be realized using multigroup training sample, and accuracy with
Predetermined speed then corresponds to the average value of accuracy and predetermined speed for this multigroup training sample, so as to further ensure accuracy
And the computational accuracy of predetermined speed.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, can also include：
Data preprocessing module is used for：Deduplication operation is carried out to training sample and test sample respectively and denoising operates, and
Training sample and test sample are stored into unified data storage.
Wherein training sample and test sample are pre-processed, can specifically be included：By training sample and test sample
In data identical (including with harmful effect or useless data) with preset noise data delete；It will training sample
Originally the data and in test sample repeated only retain portion, remaining deletion；It will be by (the i.e. data acquisition module realization of each data source
The source of data acquisition, including network, TV, newspaper etc.) in the data preparation that obtains be stored in a unified data storage
In device.So as to be realized through the above way to the data-optimized of training sample and test sample, facilitate while obtained to it
Further ensure forecasting accuracy.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, can also include：
Data structured module, is used for：The data for having identical meanings in training sample and test sample are converted into system
The training sample converted and test sample are structured as the data with unified form, and to structuring by one character
Training sample and test sample afterwards carries out quality testing, if detection acquired results meet expected results, it is determined that data are adopted
Collect successfully, otherwise, it indicates that data acquisition module resurveys data.
It should be noted that data structured model can be to be specified for certain user or with preset shadow
The operation that the training sample and test sample of the factor of sound carry out, so that the data of acquisition are more in line with actual demand；Certainly
Data structured can also be carried out to whole training samples and test sample, within protection scope of the present invention.Due to
Collected data yes-no format, including image, video, document etc., although causing to have collected these data nothings
Method effectively handles it, therefore corresponding structuring operation is carried out to it by data structured module in the application,
It ensure that being effectively treated for data.Specifically, data structured module includes data progress structuring：It will training sample
The data originally and in test sample with identical meanings are converted to corresponding character, and the data of different meanings correspond to character difference,
So as to simplify training sample and test sample；Training sample and test sample are structured as the data with unified form,
It is effectively treated with realizing；Quality testing is carried out to training sample and test sample, can be the sequential based on artificial intelligence
It is that data processing technique is realized or export training sample and test sample, by manually by the way of crowdsourcing
Realize its quality testing, it is corresponding as a result, specifically, which can be a score value, if the score value so as to obtain
Reach preset expected score value (expected results), then it is assumed that data acquire successfully, otherwise resurvey, and then ensure that
Quantity has preferable quality.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, data prediction module can be with
Including：
Data conversion unit, is used for：Predicted characteristics information is converted into the structure of extraneous setting.
During forecast function is realized, data read module can read trained deep learning model,
Using the characteristic information of the deep learning model prediction prediction time financial product, and it will predict that obtained characteristic information converts
For the structure of user setting, facilitate the understanding and reading of user.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, can also include：
Aid decision module, is used for：By the predicted characteristics information that data prediction module obtains according to the default way of output into
Row arranges, and is shown to arranging acquired results.
It can be by each specifying information (including price, behavior) in predicted characteristics information point when exporting predicted characteristics information
Other label simultaneously exports, and can also export without label but as a whole predicted characteristics information, can be by predicted characteristics information
In each specifying information different color backgrounds is respectively adopted, predicted characteristics information can also be carried on the back using identical color
Scape, etc. belongs to the way of output of predicted characteristics information, can be according to the preset way of output pair of user in the application
Prediction result is arranged；Can be that each in the displaying predicted characteristics information of one one has when showing predicted characteristics information
Body information, i.e., a certain moment only show a specifying information, whole predicted characteristics information can also be shown simultaneously, etc., from
And acquired results will be arranged and be shown, user can be facilitated to read and understand.In addition in addition to can be preset according to user
The way of output prediction result is arranged and is shown outer, it can also be arranged and is opened up in the way of system default
Show, within protection scope of the present invention.
A kind of system of artificial intelligence assessment financial product feature provided in an embodiment of the present invention, aid decision module may be used also
To include：
Feedback capture unit, is used for：Receive actual characteristic information of the financial product of extraneous input in prediction time, and profit
Incremental learning is carried out to deep learning model with actual characteristic information and corresponding test sample, after using incremental learning is carried out
Deep learning model realization characteristic information is predicted.
To input depth when the actual characteristic information of prediction time financial product and the prediction in order to realize prediction time
The data for practising model carry out incremental learning as training sample to deep learning model, and it is pre- can to improve deep learning model realization
Accuracy during survey.And incremental learning is consistent with corresponding to technical solution realization principle in the prior art, details are not described herein.
In addition, when pre- including data acquisition module 11, model training module 12, data simultaneously in the system that the application provides
Survey module 13, model chooses module 16, super ginseng preset module 17, data preprocessing module 14, data structured module 15, auxiliary
During decision-making module 18, structure diagram is as shown in Fig. 2, so as to pass through this deep learning solution end to end, Xiang Yong
Family provides financial product operation decision assistance information, and auxiliary user preferably operates in financial product, and according to user's
It feeds back (i.e. the actual characteristic information of financial product), constantly amendment and optimization algorithm, timely when market environment changes
Adaptation market operation, the suggestion for operation of optimization is provided always.And the positive negative feature of sample of financial field is all very clearly, therefore
The data analysis for realizing financial field using artificial intelligence has high accuracy.
In above-mentioned technical proposal provided in an embodiment of the present invention with to correspond to technical solution realization principle in the prior art consistent
Part and unspecified, in order to avoid excessively repeat.
The foregoing description of the disclosed embodiments enables those skilled in the art to realize or use the present invention.To this
A variety of modifications of a little embodiments will be apparent for a person skilled in the art, and the general principles defined herein can
Without departing from the spirit or scope of the present invention, to realize in other embodiments.Therefore, the present invention will not be limited
The embodiments shown herein is formed on, and is to fit to consistent with the principles and novel features disclosed herein most wide
Range.
Claims (10)
1. a kind of system of artificial intelligence assessment financial product feature, which is characterized in that including：
Data acquisition module is used for：Having before acquiring predetermined time on financial product in the characteristic information of predetermined time influences
Data and financial product predetermined time characteristic information as training sample, exist before acquisition prediction time to financial product
The characteristic information of prediction time has influential data as test sample, and the characteristic information of the financial product includes the finance
The price of product and/or behavior；
Model training module is used for：Initial deep learning model is obtained, and using the training sample to the deep learning
Model is trained, and obtains the deep learning model with forecast function；
Data prediction module, is used for：Using the test sample as the input of the deep learning model, the depth is obtained
Practise predicted characteristics information of the financial product in prediction time of model output.
2. system according to claim 1, which is characterized in that further include：
Model chooses module, is used for：The model for obtaining external world's input chooses instruction, and determines corresponding with model selection instruction
Model structure and model parameter；
Super ginseng preset module, is used for：It determines hyper parameter corresponding with the model structure, and is selected based on predetermined manner each
The value of hyper parameter obtains the initial deep learning model for including the model structure, model parameter and hyper parameter.
3. system according to claim 2, which is characterized in that the super ginseng setup module includes：
Super ginseng setting unit, for utilizing the value for emulating each hyper parameter of determining method selection.
4. system according to claim 2, which is characterized in that the model training module includes：
Model training unit, is used for：The framework for obtaining extraneous input chooses instruction, determines to choose the corresponding journey of instruction with the framework
Sequence framework generates program code corresponding with described program framework, and utilize described program based on initial deep learning model
Code is trained initial deep learning model based on the training sample.
5. system according to claim 4, which is characterized in that the model training module further includes：
Program optimization unit, is used for：Accuracy and predetermined speed of the deep learning model are obtained based on the training sample,
Judge whether the accuracy and predetermined speed reach preset requirement, if it is, determining the instruction of the deep learning model
Practice and complete, if it is not, then adjustment said program code, and indicate that the model training unit is performed and instructed using said program code
The step of practicing deep learning model, until the accuracy of the deep learning model and predetermined speed reach preset requirement.
6. system according to claim 1, which is characterized in that further include：
Data preprocessing module is used for：Deduplication operation is carried out to the training sample and the test sample respectively and denoising is grasped
Make, and the training sample and the test sample are stored into unified data storage.
7. system according to claim 6, which is characterized in that further include：
Data structured module, is used for：To there is the data conversion of identical meanings in the training sample and the test sample
For unified character, the training sample converted and test sample are structured as the data with unified form, and to knot
Training sample and test sample after structure carry out quality testing, if detection acquired results meet expected results, it is determined that number
According to acquiring successfully, otherwise, it indicates that the data acquisition module resurveys data.
8. system according to claim 1, which is characterized in that the data prediction module includes：
Data conversion unit, is used for：The predicted characteristics information is converted into the structure of extraneous setting.
9. system according to claim 1, which is characterized in that further include：
Aid decision module, is used for：By the predicted characteristics information that the data prediction module obtains according to the default way of output into
Row arranges, and is shown to arranging acquired results.
10. system according to claim 8, which is characterized in that the aid decision module further includes：
Feedback capture unit, is used for：Actual characteristic information of the financial product of extraneous input in the prediction time is received,
And incremental learning is carried out to the deep learning model using the actual characteristic information and corresponding test sample, to utilize progress
Deep learning model realization characteristic information prediction after incremental learning.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711482971.4A CN108198072A (en) | 2017-12-29 | 2017-12-29 | A kind of system of artificial intelligence assessment financial product feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711482971.4A CN108198072A (en) | 2017-12-29 | 2017-12-29 | A kind of system of artificial intelligence assessment financial product feature |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108198072A true CN108198072A (en) | 2018-06-22 |
Family
ID=62586985
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711482971.4A Pending CN108198072A (en) | 2017-12-29 | 2017-12-29 | A kind of system of artificial intelligence assessment financial product feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108198072A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109872242A (en) * | 2019-01-30 | 2019-06-11 | 北京字节跳动网络技术有限公司 | Information-pushing method and device |
| CN110175507A (en) * | 2019-04-09 | 2019-08-27 | 文远知行有限公司 | Model evaluation method, apparatus, computer equipment and storage medium |
| WO2020007177A1 (en) * | 2018-07-05 | 2020-01-09 | 京东方科技集团股份有限公司 | Quotation method executed by computer, quotation device, electronic device and storage medium |
| CN112529689A (en) * | 2020-12-16 | 2021-03-19 | 北京逸风金科软件有限公司 | Method and device for simulating bank risk pricing strategy |
| CN113127333A (en) * | 2019-12-31 | 2021-07-16 | 中移互联网有限公司 | Data processing method and device, electronic equipment and storage medium |
-
2017
- 2017-12-29 CN CN201711482971.4A patent/CN108198072A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020007177A1 (en) * | 2018-07-05 | 2020-01-09 | 京东方科技集团股份有限公司 | Quotation method executed by computer, quotation device, electronic device and storage medium |
| US11544751B2 (en) | 2018-07-05 | 2023-01-03 | Beijing Boe Technology Development Co., Ltd. | Quotation method executed by computer, quotation device, electronic device and storage medium |
| CN109872242A (en) * | 2019-01-30 | 2019-06-11 | 北京字节跳动网络技术有限公司 | Information-pushing method and device |
| CN110175507A (en) * | 2019-04-09 | 2019-08-27 | 文远知行有限公司 | Model evaluation method, apparatus, computer equipment and storage medium |
| CN110175507B (en) * | 2019-04-09 | 2023-07-04 | 文远知行有限公司 | Model evaluation method, device, computer equipment and storage medium |
| CN113127333A (en) * | 2019-12-31 | 2021-07-16 | 中移互联网有限公司 | Data processing method and device, electronic equipment and storage medium |
| CN112529689A (en) * | 2020-12-16 | 2021-03-19 | 北京逸风金科软件有限公司 | Method and device for simulating bank risk pricing strategy |
| CN112529689B (en) * | 2020-12-16 | 2024-04-26 | 北京逸风金科软件有限公司 | Simulation method and device for bank risk pricing strategy |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108198072A (en) | A kind of system of artificial intelligence assessment financial product feature | |
| CN104461855B (en) | A kind of Web automated testing method, system and device | |
| CN110084374A (en) | Construct method, apparatus and prediction technique, device based on the PU model learnt | |
| CN110348580A (en) | Construct the method, apparatus and prediction technique, device of GBDT model | |
| CN108280542A (en) | A kind of optimization method, medium and the equipment of user's portrait model | |
| CN109886928A (en) | A kind of target cell labeling method, device, storage medium and terminal device | |
| CN110263979A (en) | Method and device based on intensified learning model prediction sample label | |
| CN109978870A (en) | Method and apparatus for output information | |
| CN117494292B (en) | Engineering progress management method and system based on BIM and AI large model | |
| DE112020002684T5 (en) | A multi-process system for optimal predictive model selection | |
| CN112763967A (en) | BiGRU-based intelligent electric meter metering module fault prediction and diagnosis method | |
| Wu | MOOC learning behavior analysis and teaching intelligent decision support method based on improved decision tree C4. 5 algorithm | |
| CN116484042A (en) | Visual question-answering method combining autocorrelation and interactive guided attention mechanism | |
| CN115131700A (en) | Training method of two-way hierarchical mixed model for weakly supervised audio and video content analysis | |
| CN113159441A (en) | Prediction method and device for implementation condition of banking business project | |
| CN109101414A (en) | Based on the magnanimity UI Self -adaptive method and device thereof for burying point data | |
| CN117235527A (en) | End-to-end containerized big data model construction method, device, equipment and medium | |
| Boschetti et al. | Interactive modelling for natural resource management | |
| CN109743200B (en) | Resource feature-based cloud computing platform computing task cost prediction method and system | |
| CN108665455A (en) | The evaluation method and device of saliency prediction result | |
| CN110443282A (en) | A kind of Embryonic Stages classification method in embryo's timing image | |
| CN109711665A (en) | A kind of prediction model construction method and relevant device based on financial air control data | |
| CN107135402A (en) | A kind of method and device for recognizing TV station's icon | |
| CN113822160A (en) | Evaluation method, system and equipment of deep forgery detection model | |
| CN115511317A (en) | Wind control strategy configuration method and device |
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: 20180622 |