CN109426891A - It is a kind of high to send the forecasting system and method for turning electronic banking product - Google Patents
It is a kind of high to send the forecasting system and method for turning electronic banking product Download PDFInfo
- Publication number
- CN109426891A CN109426891A CN201810552759.9A CN201810552759A CN109426891A CN 109426891 A CN109426891 A CN 109426891A CN 201810552759 A CN201810552759 A CN 201810552759A CN 109426891 A CN109426891 A CN 109426891A
- Authority
- CN
- China
- Prior art keywords
- information
- product
- height
- training
- simple electric
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
Abstract
This application provides a kind of high send to turn the forecasting system and method for electronic banking product, wherein, the system includes: data obtaining module, for obtaining the historical transactional information of every simple electric financial product, the essential attribute information and evaluation attributes information of affiliated company and corresponding history height send and turn behavioural information;Wherein, history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;Model training module, for historical transactional information, essential attribute information and evaluation attributes information to be sent to the independent variable for turning electronic banking product prediction model as height, whether the high dependent variable for sending the record turned as model will be occurred, training height, which is sent, turns electronic banking product prediction model;Height, which is sent, turns prediction module, turns electronic banking product prediction model for sending based on trained height, and prediction target electronic financial product occurs height and send the result turned.The efficiency and accuracy rate of the application prediction are higher, and can satisfy market and send the requirement of real-time for turning electronic banking product prediction to height.
Description
Technical field
This application involves electronic banking product data digging technology fields, turn electronics gold in particular to a kind of high send
Melt the forecasting system and method for product.
Background technique
As the electronic banking product market of market economy important feature, number is cared for just from that day of birth with ten million
The heart of investor.Now there are more and more electronic banking products (such as stock, bond, insurance) for investor's choosing in the market
It selects, each listed company correspondingly provides multiple finance to investor and refer to allow investor to understand the financial situation of itself
Mark, by taking stock as an example, which can be per share capital surplus, per share undistributed profit etc..For the electronics gold of magnanimity
Melt product data, investor often tends to therefrom select the height with appreciation potential according to above-mentioned financial index and send to turn electronics
Financial product, this is mainly due to investor it has been generally acknowledged that " height, which is sent, to be turned " delivers listed company's future performance to market and will keep
The positive signal of high growth, while demand of the market to " height, which is sent, to be turned ", also can play the role of adding fuel to the flames to share price.
The relevant technologies generally use manual type and analyze above-mentioned financial index, turn electronic banking production to realize that height is sent
The prediction of product, however, the workload of artificial prediction mode be it is very huge, this directly results in height and send that turn electronic banking product pre-
The efficiency and accuracy rate of survey are lower, while being also unable to satisfy investor to the requirement of real-time of electronic banking product prediction.
Summary of the invention
In view of this, the application be designed to provide it is a kind of it is high send the forecasting system and method for turning electronic banking product,
The prediction for turning electronic banking product is sent to height to realize, the efficiency and accuracy rate of prediction are higher.
In a first aspect, the embodiment of the present application provides and a kind of high send the forecasting system for turning electronic banking product, comprising:
Data obtaining module, for obtaining the historical transactional information of every simple electric financial product, simple electric gold
Melt the essential attribute information and evaluation attributes information of the affiliated company of product, and the history corresponding to the simple electric financial product
Height, which is sent, turns behavioural information;Wherein, the history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;
Model training module is used for the historical transactional information, the essential attribute information and the evaluation attributes
Information send the independent variable for turning electronic banking product prediction model as height, send the record turned as described in the height that whether occurs
Height send the dependent variable for turning electronic banking product prediction model, and the training height, which is sent, turns electronic banking product prediction model;
Height, which is sent, turns prediction module, turns electronic banking product prediction model for sending based on the trained height, predicts mesh
Mark electronic banking product occurs height and send the result turned.
With reference to first aspect, the embodiment of the present application provides the first possible embodiment of first aspect, wherein institute
Stating essential attribute information includes one of following information or a variety of: CompanyAddress's information, company size information, the affiliated row of company
Industry information, associate's educational background information;The evaluation attributes information includes one of following information or a variety of: company is in stock
The number information that item number information that the frequency information of middle appearance, company occur in news, company occur in comment.
With reference to first aspect, the embodiment of the present application provides second of possible embodiment of first aspect, wherein institute
Stating model training module includes:
Characteristic information determination unit should for the historical transactional information of every simple electric financial product based on acquisition
The essential attribute information and evaluation attributes information of the affiliated company of simple electric financial product, determine the simple electric financial product
Characteristic information;
Model training unit, for the high note sent and turned whether to occur by corresponding using the characteristic information as independent variable
Record carries out at least one wheel model training as dependent variable;Wherein, each round model training is participated in by least two candidate families;
Prediction model generation unit, for by the combination of the smallest model of error rate in each round model training, as instruction
The height perfected, which is sent, turns electronic banking product prediction model.
The possible embodiment of second with reference to first aspect, the embodiment of the present application provide the third of first aspect
Possible embodiment, wherein the model training unit, comprising:
Model training subelement, the simple electric financial product for being used based on epicycle training determine that wheel training makes
The argument value of the characteristic information of each simple electric financial product and the dependent variable value of the simple electric financial product are right
At least two candidate families for participating in epicycle training are trained;
Model determines subelement, for according to epicycle training as a result, from at least two candidate moulds for participating in epicycle training
The minimum candidate family of error rate is determined in type;
Error product determines subelement, for determining the error sample of candidate family result mistake in epicycle training
Electronic banking product;
Weight updates subelement, for updating rule based on default weight, by the weight of simple electric financial product that malfunctions
It updates, and the weight according to the error simple electric financial product updates, updates the simple electric that the epicycle training uses
The weight of other simple electric financial products in financial product;
Stratified sampling handles subelement, right for the present weight according to updated every simple electric financial product
More simple electric financial products for participating in epicycle training carry out stratified sampling processing, and obtaining next round training needs sample to be used
This electronic banking product, into next round training, it includes that the next round training, which needs simple electric financial product to be used,
Wrong simple electric financial product.
The third possible embodiment with reference to first aspect, the embodiment of the present application provide the 4th kind of first aspect
Possible embodiment, wherein the model determines that subelement is specifically used for:
Based on the historical transactional information of every simple electric financial product, the base of the affiliated company of simple electric financial product
This attribute information and evaluation attributes information determine the argument value of the current independent variable of the correspondence of the simple electric financial product and work as
The dependent variable value of antecedents;
Successively the correspondence argument value of every simple electric financial product is inputted into epicycle respectively and completes trained at least two
A candidate family, obtains whether the simple electric financial product occurs the high prediction result sent and turned;
According to the prediction result of each sample electronic banking product and the high record for sending and turning whether occurs, from this training in rotation of participation
The minimum candidate family of error rate is determined at least two experienced candidate families.
The third possible embodiment or the 4th kind of possible embodiment with reference to first aspect, the embodiment of the present application
Provide the 5th kind of possible embodiment of first aspect, wherein the stratified sampling processing subelement is specifically used for:
From the error simple electric financial product, determine that present weight is greater than the error sample electricity of default initial weight
Sub- financial product, as first sample electronic banking product;
According to the quantity and preset quantity relationship of determining first sample electronic banking product, make from epicycle training
In other simple electric financial products in simple electric financial product, it is default to determine that the present weight of corresponding number is less than
Second simple electric financial product of initial weight;And
It needs to make using the first sample electronic banking product and the second simple electric financial product as next round training
Simple electric financial product.
The third possible embodiment or the 4th kind of possible embodiment with reference to first aspect, the embodiment of the present application
Provide the 6th kind of possible embodiment of first aspect, wherein the model training module further include:
Weight determining unit determines epicycle model training for the default value relationship according to error rate and Model Weight
The weight of the minimum candidate family of middle error rate;Wherein, the default value relationship meets that error rate is smaller, and Model Weight is got over
It is high;
The prediction model generation unit is specifically used for the smallest model of error rate in each round model training and its right
The weighted array for answering Model Weight is sent as trained height and turns electronic banking product prediction model.
Second aspect, the embodiment of the present application also provides a kind of high send to turn the prediction technique of electronic banking product, comprising:
The historical transactional information of every simple electric financial product is obtained, the base of the affiliated company of simple electric financial product
This attribute information and evaluation attributes information, and history height corresponding to the simple electric financial product send and turn behavioural information;Its
In, the history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;
The historical transactional information, the essential attribute information and the evaluation attributes information are sent as height and turns electronics
Whether the independent variable of financial product prediction model occurs high to send the record turned to send as the height to turn electronic banking product using described
The dependent variable of prediction model, the training height, which is sent, turns electronic banking product prediction model;
It is sent based on the trained height and turns electronic banking product prediction model, prediction target electronic financial product occurs high
Send the result turned.
In conjunction with second aspect, the embodiment of the present application provides the first possible embodiment of second aspect, wherein institute
Stating essential attribute information includes one of following information or a variety of: CompanyAddress's information, company size information, the affiliated row of company
Industry information, associate's educational background information;The evaluation attributes information includes one of following information or a variety of: company is in stock
The number information that item number information that the frequency information of middle appearance, company occur in news, company occur in comment.
In conjunction with second aspect, the embodiment of the present application provides second of possible embodiment of second aspect, wherein institute
It states the training height and send and turn electronic banking product prediction model, comprising:
The historical transactional information of every simple electric financial product based on acquisition, public affairs belonging to the simple electric financial product
The essential attribute information and evaluation attributes information of department, determine the characteristic information of the simple electric financial product;
Using the characteristic information as independent variable, using it is corresponding whether occur it is high send the record turned as dependent variable carry out to
Few wheel model training;Wherein, each round model training is participated in by least two candidate families;
By the combination of the smallest model of error rate in each round model training, is sent as trained height and turn electronic banking production
Product prediction model.
Height provided by the embodiments of the present application send the forecasting system and method for turning electronic banking product, and data obtaining module obtains
The historical transactional information of every simple electric financial product is taken, the essential attribute information of the affiliated company of simple electric financial product
It send with evaluation attributes information, and history height corresponding to the simple electric financial product and turns behavioural information;Wherein, the history
Height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;Model training module is by the historical transactional information, the base
This attribute information and the evaluation attributes information send the independent variable for turning electronic banking product prediction model as height, are by described
The no height that occurs send the record turned to send the dependent variable for turning electronic banking product prediction model as the height, and the training height, which is sent, turns electricity
Interest melts product prediction model;Height, which is sent, to be turned prediction module and is based on the trained height to send and turn electronic banking product prediction model,
Prediction target electronic financial product occurs height and send turning as a result, that is, it is sent by the height of training turns electronic banking product prediction
Model realization, which send the height of target electronic financial product, turns prediction, and the efficiency and accuracy rate of prediction are higher, and can satisfy city
The requirement of real-time for turning electronic banking product prediction is sent to height in field.
To enable the above objects, features, and advantages of the application to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows a kind of high structure for sending the forecasting system for turning electronic banking product provided by the embodiment of the present application and shows
It is intended to;
High model is sent in the forecasting system for turning electronic banking product to instruct Fig. 2 shows a kind of provided by the embodiment of the present application
Practice the structural schematic diagram of module;
Fig. 3, which is shown, a kind of provided by the embodiment of the present application high send in the forecasting system for turning electronic banking product model to instruct
Practice the structural schematic diagram of unit;
Fig. 4, which is shown, a kind of provided by the embodiment of the present application high send the process for turning the prediction technique of electronic banking product
Figure;
Fig. 5 shows another kind height provided by the embodiment of the present application and send the process for turning the prediction technique of electronic banking product
Figure;
Fig. 6 shows another kind height provided by the embodiment of the present application and send the process for turning the prediction technique of electronic banking product
Figure;
Fig. 7 shows another kind height provided by the embodiment of the present application and send the process for turning the prediction technique of electronic banking product
Figure;
Fig. 8 shows another kind height provided by the embodiment of the present application and send the process for turning the prediction technique of electronic banking product
Figure;
Fig. 9 shows another kind height provided by the embodiment of the present application and send the process for turning the prediction technique of electronic banking product
Figure;
Figure 10 shows a kind of structural schematic diagram of computer equipment provided by the embodiment of the present application.
Main element symbol description:
11, data obtaining module;22, model training module;33, height, which is sent, turns prediction module;221, characteristic information determines single
Member;222, model training unit;223, prediction model generation unit;224, weight determining unit;2221, model training is single
Member;2222, model determines subelement;2223, error product determines subelement;2224, weight updates subelement;2225, it is layered
Sample process subelement;1000, processor;2000, memory;3000, bus.
Specific embodiment
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
Middle attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is only
It is some embodiments of the present application, instead of all the embodiments.The application being usually described and illustrated herein in the accompanying drawings is real
The component for applying example can be arranged and be designed with a variety of different configurations.Therefore, below to the application's provided in the accompanying drawings
The detailed description of embodiment is not intended to limit claimed scope of the present application, but is merely representative of the selected reality of the application
Apply example.Based on embodiments herein, those skilled in the art institute obtained without making creative work
There are other embodiments, shall fall in the protection scope of this application.
In view of using manual type to carry out the workload right and wrong that height send the prediction for turning electronic banking product in the related technology
Often huge, this directly results in height and send the efficiency for turning electronic banking product prediction and accuracy rate lower, while being also unable to satisfy
Requirement of real-time of the investor to electronic banking product prediction.In view of this, the embodiment of the present application provide it is a kind of it is high send turn electricity
The forecasting system and method for sub- financial product turn the prediction of electronic banking product to realize to send height, the efficiency of prediction and accurate
Rate is higher.
As shown in Figure 1, sending the structural representation for turning the forecasting system of electronic banking product for height provided by the embodiments of the present application
Figure, the forecasting system include:
Data obtaining module 11, for obtaining the historical transactional information of every simple electric financial product, the simple electric
The essential attribute information and evaluation attributes information of the affiliated company of financial product, and going through corresponding to the simple electric financial product
History height, which is sent, turns behavioural information;Wherein, history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;
Model training module 22, for using historical transactional information, essential attribute information and evaluation attributes information as height
The independent variable for turning electronic banking product prediction model is sent, whether will occur high to send the record turned to send as height to turn electronic banking product
The dependent variable of prediction model, training height, which is sent, turns electronic banking product prediction model;
Height, which is sent, turns prediction module 33, turns electronic banking product prediction model for sending based on trained height, predicts target
Electronic banking product occurs height and send the result turned.
Specifically, height provided by the embodiments of the present application send the forecasting system for turning electronic banking product, pass through acquisition of information mould
Block 11 obtains the essential attribute information of historical transactional information and its affiliated company in relation to simple electric financial product and evaluation belongs to
Property information, and include whether that height, which occurs, send the history height of the record turned to send and turn behavioural information, and above- mentioned information are obtained into module
11 module transfers obtained are to model training module 22, in order to which the model training module 22 can be according to the training of the information of acquisition
Height, which is sent, turns electronic banking product prediction model, and trained height sent and turns supreme send of electronic banking product prediction model transmission and turns
Prediction module 33 send the result turned to predict, it is seen then that it is sent by the height of training so that height occurs to target electronic financial product
Turn electronic banking product prediction model realization and send a turn prediction to the height of target electronic financial product, the efficiency and accuracy rate of prediction are equal
It is higher.In addition, in the embodiment of the present application, above-mentioned height send turn electronic banking product prediction model can be it is trained in advance, this
Sample can need the arbitrary target electronic banking product predicted progress height to send a turn prediction to investor, send to meet market to height
Turn the requirement of real-time of electronic banking product prediction.
Wherein, above-mentioned electronic banking product can be stock, can also be bond, can also be insurance etc., next with
Stock carries out specific example as electronic banking product.The application obtains the historical transactional information of more sample stocks and right
The history height answered, which is sent, turns behavioural information.Above-mentioned historical transactional information and history height, which are sent, to be turned behavioural information and can be from internet net
The data-interface that (such as great wisdom, large stock trading website) accurately opens of standing is obtained, and be can also be and is climbed using network
The historical transactional information of desired acquisition and history height are sent and are changed one's profession as information crawler to local computer equipment by worm technology.
For historical transactional information, debit card, credit card etc. that the embodiment of the present application can be held by obtaining user
The pertinent transaction information of generation is determined, the relationship trading that can also be generated on the network trading platform of binding by user
Information is determined.Wherein, above-mentioned historical transactional information includes but is not limited to: per share capital surplus, per share undistributed profit, benefit
Profit increases by a year-on-year basis, stock price, the stock's coming-into market date, closing price, total market capitalisation, shareholding equity, share price rise number statistics, under stock
Fall the letter such as number statistics, advance versus decline width, net assets income ratio, basic earnings per share, net assets per share, mechanism shareholding ratio
Breath.For history height send and turns behavioural information, it can be obtained by inputting personal share code to stock public address system.
The embodiment of the present application not only obtains the historical transactional information of sample stock and corresponding history height send change one's profession for
Information also obtains the essential attribute information and evaluation attributes information of the above-mentioned affiliated company of sample stock, wherein above-mentioned company
Essential attribute information can be obtained by inputting in corresponding business information of enterprise inquiry net to inquire enterprise name, and this belongs to substantially
Property information include but is not limited to: CompanyAddress's information, company size information, the affiliated trade information of company, associate's educational background letter
Breath;Above-mentioned evaluation attributes information send the acquisition modes for turning behavioural information similar with above-mentioned historical transactional information and history height,
I.e., it is possible to obtained by data-interface and data crawler mode, unlike, which is mutual from other
Networking website, such as today's tops, Sina weibo acquisition, details are not described herein.Wherein, which includes but not
Be limited to: company's item number information that the frequency information of appearance, company occur in news in stock, company occur in comment
Number information.
It is worth noting that the embodiment of the present application can select selection in different time periods to go through according to different forecast demands
History Transaction Information, evaluation attributes information and history height, which are sent, turns behavioural information, and the embodiment of the present application not only can be using in 1 year
The information of first three quarters come predict whether to occur in the 4th season height send turn, can also be predicted not using the information of the previous year
Whether to occur for 1 year height and send to turn, at this moment, corresponding information can be chosen for different forecast demand.
It is sent in height and turns electronic banking product prediction model training stage, with the historical transactional information of above-mentioned acquisition, basic category
Property information and evaluation attributes information send the independent variable for turning electronic banking product prediction model as height to be trained, with corresponding
Height, which whether occurs, is sent the record turned as dependent variable, and training obtains height and send the parameter information for turning electronic banking product prediction model
Deng, namely obtain trained height and give to turn electronic banking product prediction model.The embodiment of the present application can be made using disaggregated model
It is sent for height and turns electronic banking product prediction model, some unknown parameter letters in model training stage i.e. train classification models
The process of breath etc..Later, so that it may sent based on above-mentioned height turn electronic banking product prediction model be target electronic financial product mention
It is serviced for prediction, only needs to produce the historical transactional information of target electronic financial product and the target electronic finance at this time
The essential attribute information and evaluation attributes information input of the affiliated company of product are sent to trained height turns electronic banking product prediction mould
In type.
In specific implementation, the embodiment of the present application can be sent based on AdaBoost classification algorithm training height turns electronic banking production
Product prediction model, that is, the embodiment of the present application is the Weak Classifier different for the training of the same training set, it is then that these are weak
Classifier gathers, and constitutes a stronger final classification device, and by continuous repetitive exercise, it is stronger to construct classification capacity
Height, which is sent, turns electronic banking product prediction model.
Height provided by the embodiments of the present application send the forecasting system for turning electronic banking product, and data obtaining module 11 obtains often
The historical transactional information of branch simple electric financial product, the essential attribute information of the affiliated company of simple electric financial product and is commented
Valence attribute information, and history height corresponding to the simple electric financial product send and turn behavioural information;Wherein, history height, which is sent, changes one's profession
Include whether that height, which occurs, send the record turned for information;Model training module 22 by historical transactional information, essential attribute information and
Evaluation attributes information is sent as height turns the independent variable of electronic banking product prediction model, using whether occur it is high send the record that turns as
Height send the dependent variable for turning electronic banking product prediction model, and training height, which is sent, turns electronic banking product prediction model;Height, which is sent, turns prediction
Module 33 is based on trained height and send and turns electronic banking product prediction model, and height occurs for prediction target electronic financial product send to turn
As a result, that is, it is sent by the height of training turns electronic banking product prediction model realization and send to the height of target electronic financial product
Turn prediction, the efficiency and accuracy rate of prediction are higher, and can satisfy market and send to height and turn the real-time of electronic banking product prediction
Property require.
Height provided by the embodiments of the present application send the forecasting system for turning electronic banking product that can also pass through data processing module
The historical transactional information of above-mentioned acquisition, history height are sent and turn behavioural information and evaluation attributes information is filtered, type turns
The processing such as change, derive, with the information that obtains that treated.Wherein, above-mentioned filtration treatment is referred to the missing letter in above- mentioned information
Breath, duplicate message etc. are filtered operation, and the above-mentioned type conversion, which can be, to be normalized, the data of separate sources
It is unified under a referential, just significant by comparison in this way, above-mentioned derivation process refers to for statistical analysis obtain
Additional information can be by spreading out to Transaction Information such as including the historical transactional information of exchange hour, transaction count
It is raw, obtain the associated statistical informations such as the average transaction count of certain stock.
The smallest mould of error rate that the embodiment of the present application is obtained at least one wheel model training by model training module 22
The group of type, which is combined into height and send, turns electronic banking product prediction model, and referring to fig. 2, above-mentioned model training module 22 specifically includes:
Characteristic information determination unit 221, for the historical transactional information of every simple electric financial product based on acquisition,
The essential attribute information and evaluation attributes information of the affiliated company of simple electric financial product, determine the simple electric financial product
Characteristic information;
Model training unit 222, for the high record for sending and turning whether to occur by corresponding using characteristic information as independent variable
At least one wheel model training is carried out as dependent variable;Wherein, each round model training is participated in by least two candidate families;
Prediction model generation unit 223, for by the combination of the smallest model of error rate in each round model training, as
Trained height, which is sent, turns electronic banking product prediction model.
Specifically, the embodiment of the present application passes through characteristic information determination unit 221 first is based on every simple electric finance production
The historical transactional information of product, the essential attribute information and evaluation attributes information of the affiliated company of simple electric financial product determine
The characteristic information of the simple electric financial product, then by model training module 22 respectively using characteristic information as independent variable,
The record turned is sent to carry out at least one wheel model training as dependent variable the corresponding height that whether occurs, it is raw finally by prediction model
At unit 223 by the combination of the smallest model of error rate in each round model training, is sent as trained height and turn electronic banking
Product prediction model.
Wherein, the characteristic information of above-mentioned simple electric financial product refers to that (such as historical transactional information is gone through to above- mentioned information
History height, which is sent, turns behavioural information and evaluation attributes information) treated as a result, features described above information can be above- mentioned information for progress
Normalization as a result, can also be feature derivative as a result, the embodiment of the present application can be adaptive based on the above- mentioned information actually obtained
The carry out Feature Selection answered.
In the embodiment of the present application for each round model training, can at least there are two candidate family participate in,
In, above-mentioned candidate family can be neural network model, SVM (Support Vector Machine, support vector machines) model,
Any combination in logistic (recurrence) model, can also be any combination of other disaggregated models.
Further, the historical transactional information of acquisition, evaluation attributes information can be more simple electric financial products
The historical trading and evaluation information of first historical time section, corresponding history height, which is sent, to be turned behavioural information and can be more simple electrics
The historical behavior information of second historical time section of financial product, and the first historical time section is earlier than the second historical time section;This
Application embodiment in training height send turn electronic banking product prediction model complete training in actual use, can be according to electronics
The historical trading and evaluation information of financial product send the height of the electronic banking product in future time and change one's profession to predict,
Therefore, when being trained, the characteristic value as characteristic information can come from historical trading and the evaluation of the first historical time section
Information, the historical behavior information for sending the characteristic value changed one's profession and be characterized to can come from the second historical time section as height.First history
Time interval between period and the second historical time section is first time interval, and height, which is sent, turns electronic banking product prediction model
Period where the historical trading and evaluation information that use when in use and needs predict whether that height, which occurs, send the future time turned
Time interval between section is the second time interval, preferably, first time interval is consistent with the second time interval.
It is worth noting that the embodiment of the present application can determine that height is sent in the following way turns electronic banking product prediction mould
Whether type reaches convergence.First way, the embodiment of the present application can send the instruction for turning electronic banking product prediction model using height
Practice the judgment mode whether wheel number reaches default exercise wheel number (such as 3 wheel), if exercise wheel number reaches preset threshold, it is determined that on
Stating height send the output for turning electronic banking product prediction model to have reached convergence, if frequency of training is not up to preset threshold, it is determined that
Not up to restrain.The second way, the embodiment of the present application can also send the output for turning electronic banking product prediction model using height
As a result and whether occur the judgement whether output error (such as error 0.0001) between the high record for sending and turning is less than default error
Mode, if output error is less than default error, it is determined that above-mentioned height send the output for turning electronic banking product prediction model to reach
To convergence, if output error is less than or equal to default error, it is determined that not up to restrain.Either above-mentioned any judgement side
The combinations of the smallest model of error rate in all wheel model trainings are sent as above-mentioned height after determination reaches convergence and turn electricity by formula
Interest melts product prediction model.
In the embodiment of the present application, as shown in figure 3, above-mentioned model training unit 222 specifically includes:
Model training subelement 2221, the simple electric financial product for being used based on epicycle training, determines the training in rotation
Practice the argument value of the characteristic information of each simple electric financial product used and the dependent variable of the simple electric financial product
Value is trained at least two candidate families for participating in epicycle training;
Model determines subelement 2222, for according to epicycle training as a result, from participate in epicycle training at least two times
The minimum candidate family of error rate is determined in modeling type;
Error product determines subelement 2223, for determining the error sample of candidate family result mistake in epicycle training
Electronic banking product;
Weight updates subelement 2224, for updating rule based on default weight, by the simple electric financial product that malfunctions
Weight updates, and the weight according to error simple electric financial product updates, and updates the simple electric finance that epicycle training uses
The weight of other simple electric financial products in product;
Stratified sampling handles subelement 2225, for the current power according to updated every simple electric financial product
Weight carries out stratified sampling processing to more simple electric financial products for participating in epicycle training, obtains next round training and needs to make
Simple electric financial product, trained into next round, next round training need simple electric financial product to be used includes
Malfunction simple electric financial product.
Specifically, each round training uses the feature of each simple electric financial product to believe in the embodiment of the present application
The dependent variable value of the argument value of breath and the simple electric financial product, at least two candidate moulds of wheel training every to above-mentioned participation
Type is trained, and based on the prediction result to whole simple electric financial products, is determined from above-mentioned at least two candidate family
The minimum candidate family of error rate weighs the error simple electric financial product for the training result mistake that candidate family determines
It updates again.
Wherein, in order to highlight error simple electric financial product, the default weight in the embodiment of the present application updates rule and refers to
Be that will the malfunction weight of simple electric financial product is turned up, the corresponding weight by correct simple electric financial product turns down.
In the specific implementation process, high in order to balance to send the predictablity rate and effect for turning electronic banking product prediction model
Rate, in the embodiment of the present application, the candidate family that every wheel training uses can be the same or different, for identical in the case of no longer
It repeats, it is special in the case of different for example: if last round of middle error rate is more than the candidate family of threshold value, next round can be with
It does not use.
Wherein, above-mentioned model determines that subelement 2222 is specifically used for: the history based on every simple electric financial product is handed over
Easy information, the essential attribute information and evaluation attributes information of the affiliated company of simple electric financial product, determines the simple electric
The argument value of the current independent variable of the correspondence of financial product and dependent variable value when antecedents;Successively by every simple electric gold
The correspondence argument value for melting product inputs at least two candidate families that epicycle completes training respectively, obtains simple electric finance
Whether product occurs the high prediction result sent and turned;According to the prediction result of each sample electronic banking product and height whether occurs send
The record turned determines the minimum candidate family of error rate from least two candidate families for participating in epicycle training.
Here, the correspondence argument value of every determining simple electric financial product is inputted this by the embodiment of the present application respectively
Wheel completes at least two candidate families of training, obtains whether the simple electric financial product occurs the high prediction result sent and turned,
According to the prediction result of each sample electronic banking product and the high record for sending and turning whether occurs, is trained at least from epicycle is participated in
The minimum candidate family of error rate is determined in two candidate families.
It is worth noting that the embodiment of the present application is after determining the minimum candidate family of error rate, it will be according to default mistake
Rate threshold value judges lowest error rate, only in lowest error rate not up to default error rate threshold, just enters next training in rotation
Practice, if lowest error rate reaches default error rate threshold (such as 0.5), training terminates to epicycle, does not continue to.
In addition, the embodiment of the present application is before entering next round training, further includes: need to make to determining next round training
Training data carries out class imbalance processing.The embodiment of the present application can carry out uneven processing using top sampling method,
Uneven processing can also be carried out using Downsapling method, can also use SMOTE (Synthetic Minority Over-
Sampling Technique) the uneven processing of method progress.Wherein, above-mentioned top sampling method refers to the too low sample of comparative example
This (that is, error simple electric financial product) duplicate sampling, so that the feature of this kind of sample is arrived by model learning;It is adopted under above-mentioned
Quadrat method refers to that the excessively high sample of comparative example (that is, correct simple electric financial product) reduces frequency in sampling, to prevent mould
The feature of this kind of pattern sheet of type overlearning;Above-mentioned SMOTE method is referred to minority class electronic banking product (that is, error
Simple electric financial product) analyze and be added in training data according to the artificial synthesized new samples of minority class sample, thus
Avoid overfitting problem.In view of the good characteristic of SMOTE method, the embodiment of the present application can carry out not according to SMOTE method
Balance Treatment.
In addition, above-mentioned stratified sampling processing subelement 2225 is specifically used for:
From error simple electric financial product, determine that present weight is greater than the error simple electric gold of default initial weight
Melt product, as first sample electronic banking product;
According to the quantity and preset quantity relationship of determining first sample electronic banking product, used from epicycle training
In other simple electric financial products in simple electric financial product, it is default initial to determine that the present weight of corresponding number is less than
Second simple electric financial product of weight;And
It is needed using first sample electronic banking product and the second simple electric financial product as next round training to be used
Simple electric financial product.
It here, first can be from all error samples after carrying out weight update to error simple electric financial product
Determine that present weight is greater than the error simple electric finance production of some or all of default initial weight of member in electronic banking product
Product, as first sample electronic banking product, which corresponds to error simple electric financial product,
Then according to the quantity of above-mentioned error simple electric financial product and error simple electric financial product and correct simple electric
Preset quantity relationship between financial product is (as the number of error simple electric financial product is produced equal to correct simple electric finance
The number of product), determine that the present weight of corresponding number is less than the second simple electric financial product of initial weight, second sample
Electronic banking product corresponds to correct simple electric financial product, by first sample electronic banking product and the second simple electric gold
Melting product as next round training needs simple electric financial product to be used.
Similarly, determine that the next round training of the next round needs simple electric gold to be used according to the prediction result of next round
Melt product and the above-mentioned prediction result according to the first round determines that next round training needs the tool of simple electric financial product to be used
Body implementation method is similar, and so on, this will not be repeated here.
Height in the embodiment of the present application send turn electronic banking product prediction model be to rely in each round model training it is wrong
The smallest model of rate is missed, as shown in Fig. 2, above-mentioned model training module 22 further include:
Weight determining unit 224 determines that epicycle model is instructed for the default value relationship according to error rate and Model Weight
The weight of the minimum candidate family of error rate in white silk;Wherein, default value relationship meets that error rate is smaller, and Model Weight is higher.
Here, default value relationship can be determined according to following formula:
Wherein, αmIndicate the determining the smallest candidate family of error rate of m wheel training
Model Weight, emIndicate the determining lowest error rate of m wheel training, β indicates that Model Weight coefficient, the Model Weight coefficient take
Being worth range is 0~1, preferably, 1/2 can be taken.
In addition, the prediction model generation unit 223 in the embodiment of the present application can also be by mistake in each round model training
The weighted array of the smallest model of rate and the Model Weight determined based on above-mentioned formula, is sent as trained height and turns electronic banking
Product prediction model.
Height in the embodiment of the present application, which is sent, to be turned prediction module 33 and send based on the obtained height of training to turn electronic banking product prediction
Model can determine that height, which occurs, send turning as a result, the height send and turns prediction module 33 and be specifically used for, and obtain for target electronic financial product
Take the historical transactional information of target electronic financial product and the essential attribute information of the affiliated company of target electronic financial product
With evaluation attributes information;By the historical transactional information of target electronic financial product, essential attribute information and evaluation attributes information
Trained height is inputted respectively and is sent and turns the smallest model of error rate that each training in rotation in electronic banking product prediction model is got, and is obtained
To the corresponding prediction result of each error rate least model;Based on the corresponding Model Weight of each error rate least model, to each
Prediction result is weighted summation, and obtain and value is determined as target electronic financial product, the high result sent and turned occurs.It can
See, using preparatory trained height to send, to turn electronic banking product prediction model can rapidly and efficiently be target electronic financial product
It carries out height and send a turn prediction, and the precision predicted is higher, the degree of automation is also higher.
Based on the same inventive concept, it is additionally provided in the embodiment of the present application and send the forecasting system for turning electronic banking product with high
Corresponding prediction technique, the principle solved the problems, such as due to the method in the embodiment of the present application and the above-mentioned prediction system of the embodiment of the present application
It unites similar, because the implementation of the method may refer to the implementation of system, overlaps will not be repeated.As shown in figure 4, real for the application
It applies height provided by example and send the flow chart for turning the prediction technique of electronic banking product, this method comprises:
S101, the historical transactional information for obtaining every simple electric financial product, public affairs belonging to the simple electric financial product
The essential attribute information and evaluation attributes information of department, and history height corresponding to the simple electric financial product send and change one's profession as letter
Breath;Wherein, history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;
S102, sent using historical transactional information, essential attribute information and evaluation attributes information as height turn electronic banking production
The independent variable of product prediction model, by whether occur it is high send the record turned as height and send turn electronic banking product prediction model because of change
Amount, training height, which is sent, turns electronic banking product prediction model;
S103, it is sent based on trained height and turns electronic banking product prediction model, prediction target electronic financial product occurs
Height send the result turned.
In specific implementation, above-mentioned essential attribute information includes one of following information or a variety of: CompanyAddress's information,
Company size information, the affiliated trade information of company, associate's educational background information;Evaluation attributes information includes one in following information
Kind is a variety of: company's item number information that the frequency information of appearance, company occur in news in stock, company go out in comment
Existing number information.
In one embodiment, as shown in figure 5, above-mentioned training height send and turns electronic banking product prediction model, comprising:
The historical transactional information of S201, every simple electric financial product based on acquisition, the simple electric financial product
The essential attribute information and evaluation attributes information of affiliated company, determine the characteristic information of the simple electric financial product;
S202, using characteristic information as independent variable, whether occur high to send the record turned as dependent variable progress using corresponding
At least one wheel model training;Wherein, each round model training is participated in by least two candidate families;
S203, by the combination of the smallest model of error rate in each round model training, sent as trained height and turn electronics
Financial product prediction model.
In another embodiment, as shown in fig. 6, the training of every wheel performs the following operations:
S301, the simple electric financial product used based on epicycle training determine each sample electricity that wheel training uses
The dependent variable value of the argument value of the characteristic information of sub- financial product and the simple electric financial product, to participation epicycle training
At least two candidate families are trained;
S302, according to epicycle training as a result, from participate in epicycle training at least two candidate families in determine error rate
Minimum candidate family;
S303, the error simple electric financial product for determining candidate family result mistake in epicycle training;
S304, rule is updated based on default weight, the weight for the simple electric financial product that malfunctions is updated, and according to error
The weight of simple electric financial product updates, and updates other simple electrics in the simple electric financial product that epicycle training uses
The weight of financial product;
S305, according to the present weight of updated every simple electric financial product, to participating in more of epicycle training
Simple electric financial product carries out stratified sampling processing, and obtaining next round training needs simple electric financial product to be used, into
Enter next round training, it includes error simple electric financial product that next round training, which needs simple electric financial product to be used,.
In yet another embodiment, as shown in fig. 7, according to epicycle training as a result, from epicycle training is participated at least
The minimum candidate family of error rate is determined in two candidate families, comprising:
S401, the historical transactional information based on every simple electric financial product, public affairs belonging to the simple electric financial product
The essential attribute information and evaluation attributes information of department, determine the independent variable of the current independent variable of the correspondence of the simple electric financial product
Value and when antecedents dependent variable value;
S402, successively by the correspondence argument value of every simple electric financial product input respectively epicycle complete training extremely
Few two candidate families, obtain whether the simple electric financial product occurs the high prediction result sent and turned;
S403, according to the prediction result of each sample electronic banking product and the high record for sending and turning whether occurs, from participation
The minimum candidate family of error rate is determined at least two candidate families of epicycle training.
In specific implementation, after determining the minimum candidate family of error rate, further includes:
According to the default value relationship of error rate and Model Weight, the candidate that error rate is minimum in epicycle model training is determined
The weight of model;Wherein, default value relationship meets that error rate is smaller, and Model Weight is higher;
By the combination of the smallest model of error rate in each round model training, is sent as trained height and turn electronic banking production
Product prediction model, specifically includes:
By the weighted array of the smallest model of error rate in each round model training and its corresponding Model Weight, as training
Good height, which is sent, turns electronic banking product prediction model.
In another embodiment, as shown in figure 8, according to the current power of updated every simple electric financial product
Weight carries out stratified sampling processing to more simple electric financial products for participating in epicycle training, obtains next round training and needs to make
Simple electric financial product, comprising:
S501, from error simple electric financial product, determine that present weight is greater than the error sample of default initial weight
Electronic banking product, as first sample electronic banking product;
S502, according to the quantity and preset quantity relationship of determining first sample electronic banking product, from epicycle training
In other simple electric financial products in the simple electric financial product used, it is pre- to determine that the present weight of corresponding number is less than
If the second simple electric financial product of initial weight;
S503, it needs to make using first sample electronic banking product and the second simple electric financial product as next round training
Simple electric financial product.
In another embodiment, turn electronic banking product prediction model as shown in figure 9, sending based on trained height,
Prediction target electronic financial product occurs height and send the result turned, comprising:
Public affairs belonging to S601, the historical transactional information for obtaining target electronic financial product and the target electronic financial product
The essential attribute information and evaluation attributes information of department;
S602, the historical transactional information of target electronic financial product, essential attribute information and evaluation attributes information are divided
Trained height is not inputted and is sent and turns the smallest model of error rate that each training in rotation in electronic banking product prediction model is got, and is obtained
The corresponding prediction result of each error rate least model;
S603, it is based on the corresponding Model Weight of each error rate least model, summation is weighted to each prediction result, and will
Obtain and value is determined as target electronic financial product and the high result sent and turned occurs.
It as shown in Figure 10, is a kind of structural schematic diagram of computer equipment provided by the embodiment of the present application, the computer
Equipment includes: processor 1000, memory 2000 and bus 3000, and the storage of memory 2000 executes instruction, when the device is running,
Communicated between processor 1000 and memory 2000 by bus 3000, processor 1000 execute stored in memory 2000 as
Under execute instruction:
The historical transactional information of every simple electric financial product is obtained, the base of the affiliated company of simple electric financial product
This attribute information and evaluation attributes information, and history height corresponding to the simple electric financial product send and turn behavioural information;Its
In, history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;
Historical transactional information, essential attribute information and evaluation attributes information are sent as height and turn electronic banking product prediction
Whether the independent variable of model will occur high to send the record turned to send as height to turn the dependent variable of electronic banking product prediction model, instruct
White silk height, which is sent, turns electronic banking product prediction model;
It is sent based on trained height and turns electronic banking product prediction model, prediction target electronic financial product generation height, which is sent, to be turned
Result.
In specific implementation, above-mentioned essential attribute information includes one of following information or a variety of: CompanyAddress's information,
Company size information, the affiliated trade information of company, associate's educational background information;Evaluation attributes information includes one in following information
Kind is a variety of: company's item number information that the frequency information of appearance, company occur in news in stock, company go out in comment
Existing number information.
In one embodiment, in the step of above-mentioned processor 1000 executes, above-mentioned training height, which is sent, turns electronic banking production
Product prediction model, comprising:
The historical transactional information of every simple electric financial product based on acquisition, public affairs belonging to the simple electric financial product
The essential attribute information and evaluation attributes information of department, determine the characteristic information of the simple electric financial product;
Using characteristic information as independent variable, the record turned is sent to carry out at least one as dependent variable the corresponding height that whether occurs
Take turns model training;Wherein, each round model training is participated in by least two candidate families;
By the combination of the smallest model of error rate in each round model training, is sent as trained height and turn electronic banking production
Product prediction model.
In another embodiment, in the step of above-mentioned processor 1000 executes, every wheel training is performed the following operations:
Based on the simple electric financial product that epicycle training uses, each simple electric finance that wheel training uses is determined
The dependent variable value of the argument value of the characteristic information of product and the simple electric financial product, at least the two of participation epicycle training
A candidate family is trained;
According to epicycle training as a result, determining that error rate is minimum from least two candidate families for participating in epicycle training
Candidate family;
Determine the error simple electric financial product of candidate family result mistake in epicycle training;
Rule is updated based on default weight, the weight for the simple electric financial product that malfunctions is updated, and according to error sample
The weight of electronic banking product updates, and updates other simple electrics finance in the simple electric financial product that epicycle training uses
The weight of product;
According to the present weight of updated every simple electric financial product, to more sample electricity for participating in epicycle training
Sub- financial product carries out stratified sampling processing, and obtaining next round training needs simple electric financial product to be used, and entrance is next
Wheel training, it includes error simple electric financial product that next round training, which needs simple electric financial product to be used,.
In yet another embodiment, in the step of above-mentioned processor 1000 executes, according to epicycle training as a result, from ginseng
The minimum candidate family of error rate is determined at least two candidate families of epicycle training, comprising:
Based on the historical transactional information of every simple electric financial product, the base of the affiliated company of simple electric financial product
This attribute information and evaluation attributes information determine the argument value of the current independent variable of the correspondence of the simple electric financial product and work as
The dependent variable value of antecedents;
Successively the correspondence argument value of every simple electric financial product is inputted into epicycle respectively and completes trained at least two
A candidate family, obtains whether the simple electric financial product occurs the high prediction result sent and turned;
According to the prediction result of each sample electronic banking product and the high record for sending and turning whether occurs, from this training in rotation of participation
The minimum candidate family of error rate is determined at least two experienced candidate families.
In specific implementation, in the step of above-mentioned processor 1000 executes, the minimum candidate family of error rate is being determined
Later, further includes:
According to the default value relationship of error rate and Model Weight, the candidate that error rate is minimum in epicycle model training is determined
The weight of model;Wherein, default value relationship meets that error rate is smaller, and Model Weight is higher;
In the step of above-mentioned processor 1000 executes, by the combination of the smallest model of error rate in each round model training,
It is sent as trained height and turns electronic banking product prediction model, specifically included:
By the weighted array of the smallest model of error rate in each round model training and its corresponding Model Weight, as training
Good height, which is sent, turns electronic banking product prediction model.
In another embodiment, above-mentioned processor 1000 execute the step of in, according to updated every sample electricity
The present weight of sub- financial product carries out stratified sampling processing to more simple electric financial products for participating in epicycle training, obtains
Simple electric financial product to be used is needed to next round training, comprising:
From error simple electric financial product, determine that present weight is greater than the error simple electric gold of default initial weight
Melt product, as first sample electronic banking product;
According to the quantity and preset quantity relationship of determining first sample electronic banking product, used from epicycle training
In other simple electric financial products in simple electric financial product, it is default initial to determine that the present weight of corresponding number is less than
Second simple electric financial product of weight;And
It is needed using first sample electronic banking product and the second simple electric financial product as next round training to be used
Simple electric financial product.
In another embodiment, above-mentioned processor 1000 execute the step of in, sent based on trained height and turn electronics
Financial product prediction model, prediction target electronic financial product occur height and send the result turned, comprising:
Obtain the historical transactional information of target electronic financial product and the base of the affiliated company of target electronic financial product
This attribute information and evaluation attributes information;
The historical transactional information of target electronic financial product, essential attribute information and evaluation attributes information are inputted respectively
Trained height, which is sent, turns the smallest model of error rate that each training in rotation in electronic banking product prediction model is got, and obtains each mistake
The corresponding prediction result of rate least model;
Based on the corresponding Model Weight of each error rate least model, summation is weighted to each prediction result, and will obtain
And value be determined as target electronic financial product and the result that height is sent turn occur.
The embodiment of the present application also provides a kind of computer readable storage medium, stored on the computer readable storage medium
There is computer program, which executes above-mentioned height and send the prediction side for turning electronic banking product when being run by processor 1000
The step of method.
Specifically, which can be general storage medium, such as mobile disk, hard disk, on the storage medium
Computer program when being run, be able to carry out above-mentioned height and send the prediction technique for turning electronic banking product, to solve current people
Efficiency brought by work prediction mode and the lower problem of accuracy, and then reach to give height and turn the pre- of electronic banking product
It surveys, the efficiency of prediction and the higher effect of accuracy.
Height provided by the embodiment of the present application send the computer program product for turning the prediction technique of electronic banking product, including
The computer readable storage medium of program code is stored, the instruction that program code includes can be used for executing previous methods embodiment
In method, specific implementation can be found in embodiment of the method, details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description
It with the specific work process of device, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
If function is realized in the form of SFU software functional unit and when sold or used as an independent product, can store
In a computer readable storage medium.Based on this understanding, the technical solution of the application is substantially in other words to existing
Having the part for the part or the technical solution that technology contributes can be embodied in the form of software products, the computer
Software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be personal meter
Calculation machine, server or network equipment etc.) execute each embodiment method of the application all or part of the steps.And it is above-mentioned
Storage medium includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory
The various media that can store program code such as (RAM, Random Access Memory), magnetic or disk.
More than, the only specific embodiment of the application, but the protection scope of the application is not limited thereto, and it is any to be familiar with
Those skilled in the art within the technical scope of the present application, can easily think of the change or the replacement, and should all cover
Within the protection scope of the application.Therefore, the protection scope of the application should be subject to the protection scope in claims.
Claims (10)
1. high the forecasting system for turning electronic banking product is sent a kind of characterized by comprising
Data obtaining module, for obtaining the historical transactional information of every simple electric financial product, which is produced
The essential attribute information and evaluation attributes information of the affiliated company of product, and history height corresponding to the simple electric financial product are sent
Turn behavioural information;Wherein, the history height, which send to change one's profession, includes whether that height, which occurs, send the record turned for information;
Model training module is used for the historical transactional information, the essential attribute information and the evaluation attributes information
The independent variable for turning electronic banking product prediction model is sent as height, send the record turned to send as the height height that whether occurs
Turn the dependent variable of electronic banking product prediction model, the training height, which is sent, turns electronic banking product prediction model;
Height, which is sent, turns prediction module, turns electronic banking product prediction model for sending based on the trained height, prediction target electricity
Sub- financial product occurs height and send the result turned.
2. system according to claim 1, which is characterized in that the essential attribute information includes one of following information
It is or a variety of: CompanyAddress's information, company size information, the affiliated trade information of company, associate's educational background information;The evaluation belongs to
Property information include one of following information or a variety of: frequency information that company occurs in the stock, company occur in news
Item number information, the number information that occurs in comment of company.
3. system according to claim 1, which is characterized in that the model training module includes:
Characteristic information determination unit, for the historical transactional information of every simple electric financial product based on acquisition, the sample
The essential attribute information and evaluation attributes information of the affiliated company of electronic banking product, determine the feature of the simple electric financial product
Information;
Model training unit, for the high record work for sending and turning whether to occur by corresponding using the characteristic information as independent variable
At least one wheel model training is carried out for dependent variable;Wherein, each round model training is participated in by least two candidate families;
Prediction model generation unit, for by the combination of the smallest model of error rate in each round model training, as training
Height send and turn electronic banking product prediction model.
4. system according to claim 3, which is characterized in that the model training unit, comprising:
Model training subelement, the simple electric financial product for being used based on epicycle training determine what wheel training used
The argument value of the characteristic information of each simple electric financial product and the dependent variable value of the simple electric financial product, to participation
At least two candidate families of epicycle training are trained;
Model determines subelement, for according to epicycle training as a result, from participate in epicycle training at least two candidate families in
Determine the minimum candidate family of error rate;
Error product determines subelement, for determining the error simple electric of candidate family result mistake in epicycle training
Financial product;
Weight updates subelement, for based on default weight update rule, the weight for the simple electric financial product that malfunctions to be updated,
And the weight according to the error simple electric financial product updates, and updates the simple electric finance that the epicycle training uses and produces
The weight of other simple electric financial products in product;
Stratified sampling handles subelement, for the present weight according to updated every simple electric financial product, to participation
More simple electric financial products of epicycle training carry out stratified sampling processing, and obtaining next round training needs sample electricity to be used
Sub- financial product, into next round training, it includes error sample that the next round training, which needs simple electric financial product to be used,
This electronic banking product.
5. system according to claim 4, which is characterized in that the model determines that subelement is specifically used for:
Based on the historical transactional information of every simple electric financial product, the basic category of the affiliated company of simple electric financial product
Property information and evaluation attributes information, determine the argument value of the current independent variable of the correspondence of the simple electric financial product and work as cause
The dependent variable value of variable;
The correspondence argument value of every simple electric financial product is successively inputted at least two times that epicycle completes training respectively
Modeling type, obtains whether the simple electric financial product occurs the high prediction result sent and turned;
According to the prediction result of each sample electronic banking product and the high record for sending and turning whether occurs, from participation epicycle training
The minimum candidate family of error rate is determined at least two candidate families.
6. system according to claim 4 or 5, which is characterized in that the stratified sampling processing subelement is specifically used for:
From the error simple electric financial product, determine that present weight is greater than the error simple electric gold of default initial weight
Melt product, as first sample electronic banking product;
According to the quantity and preset quantity relationship of determining first sample electronic banking product, used from epicycle training
In other simple electric financial products in simple electric financial product, it is default initial to determine that the present weight of corresponding number is less than
Second simple electric financial product of weight;And
It is needed using the first sample electronic banking product and the second simple electric financial product as next round training to be used
Simple electric financial product.
7. system according to claim 4 or 5, which is characterized in that the model training module further include:
Weight determining unit determines wrong in epicycle model training for the default value relationship according to error rate and Model Weight
The accidentally weight of the minimum candidate family of rate;Wherein, the default value relationship meets that error rate is smaller, and Model Weight is higher;
The prediction model generation unit is specifically used for the smallest model of error rate in each round model training and its corresponding mould
The weighted array of type weight is sent as trained height and turns electronic banking product prediction model.
8. high the prediction technique for turning electronic banking product is sent a kind of characterized by comprising
The historical transactional information of every simple electric financial product is obtained, the basic category of the affiliated company of simple electric financial product
Property information and evaluation attributes information, and history height corresponding to the simple electric financial product send and turn behavioural information;Wherein, institute
State history height send change one's profession for information include whether occur height send turn record;
The historical transactional information, the essential attribute information and the evaluation attributes information are sent as height and turns electronic banking
Whether the independent variable of product prediction model occurs high to send the record turned to send as the height to turn electronic banking product prediction using described
The dependent variable of model, the training height, which is sent, turns electronic banking product prediction model;
It is sent based on the trained height and turns electronic banking product prediction model, prediction target electronic financial product generation height, which is sent, to be turned
Result.
9. according to the method described in claim 8, it is characterized in that, the essential attribute information includes one of following information
It is or a variety of: CompanyAddress's information, company size information, the affiliated trade information of company, associate's educational background information;The evaluation belongs to
Property information include one of following information or a variety of: frequency information that company occurs in the stock, company occur in news
Item number information, the number information that occurs in comment of company.
10. according to the method described in claim 8, it is characterized in that, the training height send and turns electronic banking product prediction
Model, comprising:
The historical transactional information of every simple electric financial product based on acquisition, the affiliated company of simple electric financial product
Essential attribute information and evaluation attributes information, determine the characteristic information of the simple electric financial product;
Using the characteristic information as independent variable, the record turned is sent to carry out at least one as dependent variable the corresponding height that whether occurs
Take turns model training;Wherein, each round model training is participated in by least two candidate families;
By the combination of the smallest model of error rate in each round model training, send that turn electronic banking product pre- as trained height
Survey model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810552759.9A CN109426891A (en) | 2018-05-31 | 2018-05-31 | It is a kind of high to send the forecasting system and method for turning electronic banking product |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810552759.9A CN109426891A (en) | 2018-05-31 | 2018-05-31 | It is a kind of high to send the forecasting system and method for turning electronic banking product |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109426891A true CN109426891A (en) | 2019-03-05 |
Family
ID=65514488
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810552759.9A Pending CN109426891A (en) | 2018-05-31 | 2018-05-31 | It is a kind of high to send the forecasting system and method for turning electronic banking product |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109426891A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110415119A (en) * | 2019-07-30 | 2019-11-05 | 中国工商银行股份有限公司 | Model training, bill business prediction technique, device, storage medium and equipment |
CN110766465A (en) * | 2019-10-24 | 2020-02-07 | 开鑫金融科技服务江苏有限公司 | Financial product evaluation method and verification method and device thereof |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7930195B2 (en) * | 2002-10-11 | 2011-04-19 | Accenture Global Services Limited | Strategic management and budgeting tools |
CN105930934A (en) * | 2016-04-27 | 2016-09-07 | 北京物思创想科技有限公司 | Prediction model demonstration method and device and prediction model adjustment method and device |
CN108053314A (en) * | 2018-01-31 | 2018-05-18 | 国信优易数据有限公司 | A kind of Loan Demand Forecasting Methodology |
-
2018
- 2018-05-31 CN CN201810552759.9A patent/CN109426891A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7930195B2 (en) * | 2002-10-11 | 2011-04-19 | Accenture Global Services Limited | Strategic management and budgeting tools |
CN105930934A (en) * | 2016-04-27 | 2016-09-07 | 北京物思创想科技有限公司 | Prediction model demonstration method and device and prediction model adjustment method and device |
CN108053314A (en) * | 2018-01-31 | 2018-05-18 | 国信优易数据有限公司 | A kind of Loan Demand Forecasting Methodology |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110415119A (en) * | 2019-07-30 | 2019-11-05 | 中国工商银行股份有限公司 | Model training, bill business prediction technique, device, storage medium and equipment |
CN110415119B (en) * | 2019-07-30 | 2022-03-25 | 中国工商银行股份有限公司 | Model training method, bill transaction prediction method, model training device, bill transaction prediction device, storage medium and equipment |
CN110766465A (en) * | 2019-10-24 | 2020-02-07 | 开鑫金融科技服务江苏有限公司 | Financial product evaluation method and verification method and device thereof |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | FinRL: A deep reinforcement learning library for automated stock trading in quantitative finance | |
Ketter et al. | A multiagent competitive gaming platform to address societal challenges | |
Gatti et al. | Macroeconomics from the Bottom-up | |
Farmer et al. | An ecological perspective on the future of computer trading | |
Shi et al. | An adaptive estimation of distribution algorithm for multipolicy insurance investment planning | |
Brabazon et al. | An introduction to evolutionary computation in finance | |
Mironov et al. | Structural changes and economic growth in the world economy and Russia | |
Schulenburg et al. | Explorations in LCS models of stock trading | |
Hsu | An integrated portfolio optimisation procedure based on data envelopment analysis, artificial bee colony algorithm and genetic programming | |
Zou et al. | Does artificial intelligence promote industrial upgrading? Evidence from China | |
Kannan et al. | A novel software package selection method using teaching–learning based optimization and multiple criteria decision making | |
Wang et al. | An experimental study on real-options strategies | |
CN109426891A (en) | It is a kind of high to send the forecasting system and method for turning electronic banking product | |
Kinsella | Words to the wise: Stock flow consistent modeling of financial instability | |
CN110322149A (en) | One-to-many bipartite matching method in time bank based on multiple index evaluation | |
CN107239853B (en) | Intelligent housekeeper system based on cloud computing and working method thereof | |
Carr et al. | Generalized compounding and growth optimal portfolios: Reconciling kelly and samuelson | |
Booth | Automated algorithmic trading: Machine learning and agent-based modelling in complex adaptive financial markets | |
CN107844874A (en) | Enterprise operation problem analysis system and its method | |
CN112927040A (en) | Intelligent recommendation method for financial service platform | |
Hasheminejad et al. | Syndicated venture capital portfolio companies selection: a fuzzy inference system–agent-based approach | |
Huang | Volatility forecasting by quantile regression | |
Mehlawat et al. | An integrated fuzzy-grey relational analysis approach to portfolio optimization | |
Mahjoub et al. | A hybrid model for customer credit scoring in stock brokerages using data mining approach | |
Tziralis et al. | Prediction markets: an information aggregation perspective to the forecasting problem |
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 | ||
CB02 | Change of applicant information |
Address after: 101-8, 1st floor, building 31, area 1, 188 South Fourth Ring Road West, Fengtai District, Beijing Applicant after: Guoxin Youyi Data Co., Ltd Address before: 100070, No. 188, building 31, headquarters square, South Fourth Ring Road West, Fengtai District, Beijing Applicant before: SIC YOUE DATA Co.,Ltd. |
|
CB02 | Change of applicant information | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190305 |
|
RJ01 | Rejection of invention patent application after publication |