CN108446890A - A kind of examination & approval model training method, computer readable storage medium and terminal device - Google Patents
A kind of examination & approval model training method, computer readable storage medium and terminal device Download PDFInfo
- Publication number
- CN108446890A CN108446890A CN201810161032.8A CN201810161032A CN108446890A CN 108446890 A CN108446890 A CN 108446890A CN 201810161032 A CN201810161032 A CN 201810161032A CN 108446890 A CN108446890 A CN 108446890A
- Authority
- CN
- China
- Prior art keywords
- examination
- approval
- sample
- matrix
- sample matrix
- 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
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
Abstract
The invention belongs to a kind of field of computer technology more particularly to examination & approval model training method, computer readable storage medium and terminal devices.The method chooses the examination & approval sample of preset number from history approval record;Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, the examination & approval sample to be quantized;By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the covariance matrix of the examination & approval sample matrix;The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and the maximum characteristic value of numerical value of preset number is chosen as dominant eigenvalue from the characteristic value;The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified;Preset examination & approval model is trained using the examination & approval sample matrix after the simplification.Under the premise of ensureing precision, the complexity being trained to examination & approval model is substantially reduced, a large amount of training time is saved.
Description
Technical field
The invention belongs to field of computer technology more particularly to a kind of examination & approval model training method, computer-readable storages
Medium and terminal device.
Background technology
During carrying out various examination & approval, such as in the approval process without mortgage deduction and exemption, often by manually to each
Information in a examination & approval dimension is considered and is determined final approval results, examination & approval efficiency it is low and it is easy by it is artificial because
Element interference.
It is widely applied very much in recent years, nerual network technique has in every field, neural network is by a large amount of, simple
Single processing unit widely interconnects and the complex networks system that is formed, it reflects many substantially special of human brain function
Sign, is a highly complex non-linear dynamic learning system.Neural network has large-scale parallel, distributed storage and place
Reason, self-organizing, adaptive and self-learning ability, be particularly suitable for processing need to consider simultaneously many factors and condition, inaccurately and
Fuzzy information-processing problem.
Therefore, become a preferable selection using nerual network technique to carry out various examination and approval works, but right
When neural network model is trained, if the examination & approval dimension of the examination and approval work is more, trained complicated journey can be greatly increased
Degree, expends a large amount of training time.
Invention content
In view of this, an embodiment of the present invention provides a kind of examination & approval model training method, computer readable storage medium and
Terminal device can greatly increase trained complexity to solve when the examination & approval dimension of examination and approval work is more, expend a large amount of
The problem of training time.
The first aspect of the embodiment of the present invention provides a kind of examination & approval model training method, may include:
The examination & approval sample of preset number is chosen from history approval record, the examination & approval sample includes that approval results are to pass through
Positive sample and negative sample that approval results are refusal, and the ratio between the number of the positive sample and the number of the negative sample are pre-
If ratio range in;
Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, the examination & approval sample to be quantized
This;
By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the covariance of the examination & approval sample matrix
Matrix, wherein the arbitrary data line of the examination & approval sample matrix is corresponding with the examination & approval sample that one quantizes;
The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and preset number is chosen from the characteristic value
The maximum characteristic value of numerical value as dominant eigenvalue;
The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified, after the simplification
It examines and only retains row corresponding with the dominant eigenvalue in sample matrix;
Preset examination & approval model is trained using the examination & approval sample matrix after the simplification, obtains trained examination & approval
Model.
The second aspect of the embodiment of the present invention provides a kind of computer readable storage medium, the computer-readable storage
Media storage has computer-readable instruction, the computer-readable instruction to realize following steps when being executed by processor:
The examination & approval sample of preset number is chosen from history approval record, the examination & approval sample includes that approval results are to pass through
Positive sample and negative sample that approval results are refusal, and the ratio between the number of the positive sample and the number of the negative sample are pre-
If ratio range in;
Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, the examination & approval sample to be quantized
This;
By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the covariance of the examination & approval sample matrix
Matrix, wherein the arbitrary data line of the examination & approval sample matrix is corresponding with the examination & approval sample that one quantizes;
The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and preset number is chosen from the characteristic value
The maximum characteristic value of numerical value as dominant eigenvalue;
The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified, after the simplification
It examines and only retains row corresponding with the dominant eigenvalue in sample matrix;
Preset examination & approval model is trained using the examination & approval sample matrix after the simplification, obtains trained examination & approval
Model.
The third aspect of the embodiment of the present invention provides a kind of examination & approval model training terminal device, including memory, processing
Device and it is stored in the computer-readable instruction that can be run in the memory and on the processor, the processor executes
Following steps are realized when the computer-readable instruction:
The examination & approval sample of preset number is chosen from history approval record, the examination & approval sample includes that approval results are to pass through
Positive sample and negative sample that approval results are refusal, and the ratio between the number of the positive sample and the number of the negative sample are pre-
If ratio range in;
Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, the examination & approval sample to be quantized
This;
By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the covariance of the examination & approval sample matrix
Matrix, wherein the arbitrary data line of the examination & approval sample matrix is corresponding with the examination & approval sample that one quantizes;
The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and preset number is chosen from the characteristic value
The maximum characteristic value of numerical value as dominant eigenvalue;
The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified, after the simplification
It examines and only retains row corresponding with the dominant eigenvalue in sample matrix;
Preset examination & approval model is trained using the examination & approval sample matrix after the simplification, obtains trained examination & approval
Model.
Existing advantageous effect is the embodiment of the present invention compared with prior art:The embodiment of the present invention is from history approval record
The middle examination & approval sample for choosing preset number;Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension,
The examination & approval sample to be quantized;By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the examination & approval sample
The covariance matrix of this matrix;The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and from the characteristic value
The maximum characteristic value of numerical value of preset number is chosen as dominant eigenvalue;The examination & approval sample matrix is carried out to simplify processing, is obtained
Examination & approval sample matrix after to simplification;Preset examination & approval model is trained using the examination & approval sample matrix after the simplification,
Obtain trained examination & approval model.Through the invention, be not directly using original examination & approval sample to preset examination & approval model into
Row training, but the examination & approval sample of numeralization is formed into examination & approval sample matrix, and calculate the covariance of the examination & approval sample matrix
The characteristic value of matrix therefrom chooses the maximum characteristic value of numerical value of preset number as dominant eigenvalue, reservation and dominant eigenvalue
Information in corresponding examination & approval dimension is utmostly ensureing training result essence without retaining the information in other examination & approval dimensions
Under the premise of degree, the complexity being trained to examination & approval model is substantially reduced, a large amount of training time is saved.
Description of the drawings
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to embodiment or description of the prior art
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description be only the present invention some
Embodiment for those of ordinary skill in the art without having to pay creative labor, can also be according to these
Attached drawing obtains other attached drawings.
Fig. 1 is a kind of one embodiment flow chart of examination & approval model training method in the embodiment of the present invention;
Fig. 2 is the correspondence figure examined between sample matrix and examination & approval sample;
Fig. 3 is the examination & approval sample matrix after simplifying and the correspondence figure between dominant eigenvalue;
Fig. 4 is a kind of one embodiment structure chart of examination & approval model training apparatus in the embodiment of the present invention;
Fig. 5 is a kind of schematic block diagram of examination & approval model training terminal device in the embodiment of the present invention.
Specific implementation mode
In order to make the invention's purpose, features and advantages of the invention more obvious and easy to understand, below in conjunction with the present invention
Attached drawing in embodiment, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that disclosed below
Embodiment be only a part of the embodiment of the present invention, and not all embodiment.Based on the embodiments of the present invention, this field
All other embodiment that those of ordinary skill is obtained without making creative work, belongs to protection of the present invention
Range.
Referring to Fig. 1, in the embodiment of the present invention it is a kind of examination & approval model training method one embodiment may include:
Step S101, the examination & approval sample of preset number is chosen from history approval record.
Each examination & approval sample may include following information:People's row reference information, social security information, common reserve fund information, educational background letter
Breath, Police Information, occupational information, archive information and approval results.Wherein, approval results are the output information for examining model,
Other information is the input information for examining model.
The examination & approval sample include approval results be by positive sample and negative sample that approval results are refusal, and it is described
The ratio between the number of positive sample and the number of the negative sample are in preset ratio range.
Distinguishingly, in the examination & approval sample of selection, positive sample can be kept consistent with negative sample number.If for example, choosing altogether
10000 examination & approval samples are taken, then wherein each 5000 of positive sample and negative sample, ensure the balance of training result with this.
It is possible to further be divided into several amount sections, each amount according to the amount of approval results to positive sample
The sample number in section is consistent.
Specifically, first examination & approval amount is normalized, the value being translated into [0,1] section, conversion is public
Formula is:S′n=Sn/Smax, for a certain sample of serial number n, examination & approval amount is Sn, SmaxFor the desirable maximum of examination & approval amount
Numerical value, S 'nFor the examination & approval amount after normalization.Distinguishingly, for negative sample, it is believed that it is 0 that it, which examines amount,.
To the examination & approval amount after normalization, it is classified as N number of segment, by taking N=5 as an example, respectively (0,0.2], (0.2,
0.4], (0.4,0.6], (0.6,0.8], (0.8,1], if positive sample totally 5000, the sample of each segment is 1000.
By the above process, the sample of approval results equiblibrium mass distribution has been selected, ensure that the examination & approval model pair trained
In the sample of each segment be all applicable.
Step S102, numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, obtains numerical value
The examination & approval sample of change.
Specifically, for examining people's row reference information in sample, record of bad behavior number therein is counted, and be arranged not
The threshold value of good record, if being more than the threshold value, people's row reference information value turns to 1, if being less than the threshold value, calculates record of bad behavior
The ratio of number and the threshold value, using the ratio as numeralization treated result.Distinguishingly, if record of bad behavior number is 0,
People's row reference information value turns to 0.
Again for examining the social security information in sample, successive tranche social security months therein and current total value are counted,
And months threshold value and sum threshold are set, and if successive tranche social security months are more than months threshold value, the first numerical value of social security information
It is 1, if being less than the threshold value, calculates the ratio of successive tranche social security months and the threshold value, using the ratio as the first numerical value, together
Reason calculates the second value of social security information according to current total value and sum threshold, finally seeks the average value of the two, believes as social security
Breath numeralization treated result.
It is handled by the above numeralization, information of the examination & approval sample in each examination & approval dimension is converted to [0,1]
Value in section, convenient for being subsequently trained to examination & approval model.
Step S103, the examination & approval sample of the numeralization is formed into examination & approval sample matrix, and calculates the examination & approval sample moment
The covariance matrix of battle array.
It is possible, firstly, to which the examination & approval sample of the numeralization is formed following examination & approval sample matrix:
As shown in Fig. 2, the arbitrary data line of the examination & approval sample matrix is corresponding with the examination & approval sample that one quantizes.
Wherein, X is the examination & approval sample matrix, xijThe examination & approval sample to quantize for i-th is examined at j-th in dimension
Information, 1≤i≤n, 1≤j≤p, n are the sum of the examination & approval sample of the numeralization, and p is the number of the examination & approval dimension.
Then, the covariance matrix of the examination & approval sample matrix is calculated according to the following formula:
Wherein, R is the covariance matrix of the examination & approval sample matrix, 1≤a≤p, 1≤b≤p.
Step S104, the characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and is selected from the characteristic value
Take the maximum characteristic value of the numerical value of preset number as dominant eigenvalue;
First, characteristic equation is solved | λ I-R |=0, find out eigenvalue λa, wherein I is unit matrix, 1≤a≤p.
Then, the contribution rate of each characteristic value is calculated according to the following formula:
Wherein, ηaIt is characterized value λaContribution rate.
Finally, using the maximum preceding m characteristic value of the numerical value for meeting following condition as dominant eigenvalue:
And
Wherein ηthresholdFor preset contribution rate threshold value.
Step S105, the examination & approval sample matrix is carried out simplifying processing, the examination & approval sample matrix after being simplified.
Examination & approval sample matrix after the simplification can be expressed as:
As shown in figure 3, only retain row corresponding with the dominant eigenvalue in examination & approval sample matrix after the simplification, and
Delete other row.
Step S106, preset examination & approval model is trained using the examination & approval sample matrix after the simplification, is instructed
The examination & approval model perfected.
Specifically, a wheel is carried out to the examination & approval model using the examination & approval sample matrix after the simplification first to train, and
The global error of epicycle training is calculated according to the following formula:
Wherein, EtFor the training error of t-th of training sample, ztFor the training output valve of t-th of training sample, ctFor t
The theoretical output valve of a training sample, t-th of training sample are the t line numbers of the examination & approval sample matrix after the simplification
According to 1≤t≤n.
If the global error is more than preset error threshold, the examination & approval model is adjusted, is then back to and holds
The row examination & approval sample matrix using after the simplification carries out the examination & approval model step of one wheel training, until described complete
Until office's error is less than the error threshold.
If the global error is less than the error threshold, current examination & approval model is determined as described trained careful
Criticize model.
After training the examination & approval model, the examination & approval model may be used, pending application is examined.First
Pending application is subjected to numeralization processing (with the content in step S102, details are not described herein again), is then input to described
It examines in model, and obtains output valve.If output valve is 0, refuse this application;If output valve is not 0, this application is agreed to,
Then specific amount is calculated, i.e., is multiplied by S with output valvemax, you can obtain actual examination & approval amount.
In conclusion the embodiment of the present invention chooses the examination & approval sample of preset number from history approval record;It examines described
Information of the lot sample sheet in each examination & approval dimension carries out numeralization processing, the examination & approval sample to be quantized;By the numeralization
Examination & approval sample composition examination & approval sample matrix, and calculate it is described examination & approval sample matrix covariance matrix;Calculate the examination & approval sample
The characteristic value of the covariance matrix of this matrix, and from the characteristic value choose preset number the maximum characteristic value conduct of numerical value
Dominant eigenvalue;The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified;Use the simplification
Examination & approval sample matrix afterwards is trained preset examination & approval model, obtains trained examination & approval model.Through the invention, it is not
Directly preset examination & approval model is trained using original examination & approval sample, but the examination & approval sample of numeralization is formed and is examined
Sample matrix, and the characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, therefrom choose the numerical value of preset number most
Big characteristic value only retains the information in examination & approval dimension corresponding with dominant eigenvalue, without retaining other examine as dominant eigenvalue
The information in dimension is criticized, under the premise of utmostly ensureing training result precision, substantially reduces and examination & approval model is instructed
Experienced complexity saves a large amount of training time.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process
Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit
It is fixed.
Corresponding to a kind of examination & approval model training method described in foregoing embodiments, Fig. 4 shows that the embodiment of the present invention provides
A kind of examination & approval model training apparatus one embodiment structure chart.
In the present embodiment, a kind of examination & approval model training apparatus may include:
It examines sample and chooses module 401, the examination & approval sample for choosing preset number from history approval record is described to examine
Lot sample originally include approval results be by positive sample and approval results be refusal negative sample, and the number of the positive sample with
The ratio between number of the negative sample is in preset ratio range;
Quantize processing module 402, for quantizing to information of the examination & approval sample in each examination & approval dimension
Processing, the examination & approval sample to be quantized;
Covariance matrix computing module 403 for the examination & approval sample composition of the numeralization to be examined sample matrix, and is counted
Calculate it is described examination & approval sample matrix covariance matrix, wherein it is described examination & approval sample matrix arbitrary data line with a number
The examination & approval sample of value corresponds to;
Dominant eigenvalue chooses module 404, the characteristic value of the covariance matrix for calculating the examination & approval sample matrix, and from
The maximum characteristic value of numerical value of preset number is chosen in the characteristic value as dominant eigenvalue;
Simplify processing module 405, simplifies processing, the examination & approval sample after being simplified for being carried out to the examination & approval sample matrix
This matrix only retains row corresponding with the dominant eigenvalue in the examination & approval sample matrix after the simplification;
Model training module 406, for being carried out to preset examination & approval model using the examination & approval sample matrix after the simplification
Training, obtains trained examination & approval model.
Further, the covariance matrix computing module may include:
Sample matrix component units, for the examination & approval sample of the numeralization to be formed following examination & approval sample matrix:
Wherein, X is the examination & approval sample matrix, xijThe examination & approval sample to quantize for i-th is examined at j-th in dimension
Information, 1≤i≤n, 1≤j≤p, n are the sum of the examination & approval sample of the numeralization, and p is the number of the examination & approval dimension;
Covariance matrix computing unit, the covariance matrix for calculating the examination & approval sample matrix according to the following formula:
Wherein, R is the covariance matrix of the examination & approval sample matrix, 1≤a≤p, 1≤b≤p.
Further, the dominant eigenvalue extraction module may include:
Characteristic value computing unit, for solving characteristic equation | λ I-R |=0, find out eigenvalue λa, wherein I is unit matrix,
1≤a≤p;
Contribution rate computing unit, the contribution rate for calculating each characteristic value according to the following formula:
Wherein, η a are characterized value λaContribution rate;
Dominant eigenvalue selection unit, the maximum preceding m characteristic value of numerical value for that will meet following condition is as main feature
Value:
AndWherein ηthresholdFor preset contribution rate threshold value.
Further, the model training module may include:
Global error computing unit, for carrying out one to the examination & approval model using the examination & approval sample matrix after the simplification
Wheel training, and the global error of epicycle training is calculated according to the following formula:
Wherein, EtFor the training error of t-th of training sample, ztFor the training output valve of t-th of training sample, ctFor t
The theoretical output valve of a training sample, t-th of training sample are the t line numbers of the examination & approval sample matrix after the simplification
According to 1≤t≤n;
Model adjustment unit, if for the global error be more than preset error threshold, to the examination & approval model into
Row adjustment;
Model determination unit, it is if being less than the error threshold for the global error, current examination & approval model is true
It is set to the trained examination & approval model.
It is apparent to those skilled in the art that for convenience and simplicity of description, the device of foregoing description,
The specific work process of module and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, is not described in detail or remembers in some embodiment
The part of load may refer to the associated description of other embodiments.
Fig. 5 shows a kind of schematic block diagram of examination & approval model training terminal device provided in an embodiment of the present invention, in order to just
In explanation, illustrate only and the relevant part of the embodiment of the present invention.
In the present embodiment, the examination & approval model training terminal device 5 can be desktop PC, notebook, palm
The computing devices such as computer and cloud server.The examination & approval model training terminal device 5 may include:Processor 50, memory 51 with
And it is stored in the computer-readable instruction 52 that can be run in the memory 51 and on the processor 50, such as execute above-mentioned
Examination & approval model training method computer-readable instruction.The processor 50 is realized when executing the computer-readable instruction 52
Step in above-mentioned each examination & approval model training method embodiment, such as step S101 to S106 shown in FIG. 1.Alternatively, described
Processor 50 realizes the function of each module/unit in above-mentioned each device embodiment, example when executing the computer-readable instruction 52
The function of module 401 to 406 as shown in Figure 4.
Illustratively, the computer-readable instruction 52 can be divided into one or more module/units, one
Or multiple module/units are stored in the memory 51, and executed by the processor 50, to complete the present invention.Institute
It can be the series of computation machine readable instruction section that can complete specific function, the instruction segment to state one or more module/units
For describing implementation procedure of the computer-readable instruction 52 in the examination & approval model training terminal device 5.
The processor 50 can be central processing unit (Central Processing Unit, CPU), can also be
Other general processors, digital signal processor (Digital Signal Processor, DSP), application-specific integrated circuit
(Application Specific Integrated Circuit, ASIC), field programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor can also be any conventional processor
Deng.
The memory 51 can be the internal storage unit of the examination & approval model training terminal device 5, such as examine mould
The hard disk or memory of type training terminal device 5.The memory 51 can also be the outer of the examination & approval model training terminal device 5
The plug-in type hard disk being equipped in portion's storage device, such as the examination & approval model training terminal device 5, intelligent memory card (Smart
Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card) etc..Further,
The memory 51 can also be both including the internal storage unit for examining model training terminal device 5 or including external storage
Equipment.The memory 51 is for storing needed for the computer-readable instruction and the examination & approval model training terminal device 5
Other instruction and datas.The memory 51 can be also used for temporarily storing the data that has exported or will export.
Each functional unit in each embodiment of the present invention can be integrated in a processing unit, can also be each
Unit physically exists alone, can also be during two or more units are integrated in one unit.Above-mentioned integrated unit both may be used
It realizes, can also be realized in the form of SFU software functional unit in the form of using hardware.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can be stored in a computer read/write memory medium.Based on this understanding, technical scheme of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
Embody, which is stored in a storage medium, including several computer-readable instructions use so that
One computer equipment (can be personal computer, server or the network equipment etc.) executes each embodiment institute of the present invention
State all or part of step of method.And storage medium above-mentioned includes:USB flash disk, mobile hard disk, read-only memory (ROM, Read-
Only Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can be with
Store the medium of computer-readable instruction.
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations;Although with reference to aforementioned reality
Applying example, invention is explained in detail, it will be understood by those of ordinary skill in the art that:It still can be to aforementioned each
Technical solution recorded in embodiment is modified or equivalent replacement of some of the technical features;And these are changed
Or it replaces, the spirit and scope for various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution.
Claims (10)
1. a kind of examination & approval model training method, which is characterized in that including:
The examination & approval sample of preset number is chosen from history approval record, the examination & approval sample includes that approval results are by just
Sample and approval results are the negative sample of refusal, and the ratio between the number of the positive sample and the number of the negative sample are preset
In ratio range;
Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, the examination & approval sample to be quantized;
By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the covariance square of the examination & approval sample matrix
Battle array, wherein the arbitrary data line of the examination & approval sample matrix is corresponding with the examination & approval sample that one quantizes;
The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and chooses the number of preset number from the characteristic value
It is worth maximum characteristic value as dominant eigenvalue;
The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified, the examination & approval after the simplification
Only retain row corresponding with the dominant eigenvalue in sample matrix;
Preset examination & approval model is trained using the examination & approval sample matrix after the simplification, obtains trained examination & approval mould
Type.
2. examination & approval model training method according to claim 1, which is characterized in that the examination & approval sample by the numeralization
This composition examines sample matrix:
The examination & approval sample of the numeralization is formed into following examination & approval sample matrix:
Wherein, X is the examination & approval sample matrix, xijThe examination & approval sample to quantize for i-th examines the letter in dimension at j-th
Breath, 1≤i≤n, 1≤j≤p, n are the sum of the examination & approval sample of the numeralization, and p is the number of the examination & approval dimension.
3. examination & approval model training method according to claim 2, which is characterized in that described to calculate the examination & approval sample matrix
Covariance matrix include:
The covariance matrix of the examination & approval sample matrix is calculated according to the following formula:
Wherein, R is the covariance matrix of the examination & approval sample matrix, 1≤a≤p, 1≤b≤p.
4. examination & approval model training method according to claim 3, which is characterized in that described to calculate the examination & approval sample matrix
Covariance matrix characteristic value, and from the characteristic value choose preset number the maximum characteristic value of numerical value as main feature
Value includes:
Solve characteristic equation | λ I-R |=0, find out eigenvalue λa, wherein I is unit matrix, 1≤a≤p;
The contribution rate of each characteristic value is calculated according to the following formula:
Wherein, ηaIt is characterized value λaContribution rate;
Using the maximum preceding m characteristic value of the numerical value for meeting following condition as dominant eigenvalue:
AndWherein ηthresholdFor preset contribution rate threshold value.
5. examination & approval model training method according to any one of claim 1 to 4, which is characterized in that described in the use
Examination & approval sample matrix after simplification is trained preset examination & approval model, obtains trained examination & approval model and includes:
A wheel is carried out to the examination & approval model to train, and calculate epicycle according to the following formula using the examination & approval sample matrix after the simplification
Trained global error:
Wherein, EtFor the training error of t-th of training sample, ztFor the training output valve of t-th of training sample, ctIt is instructed for t-th
Practice sample theoretical output valve, t-th of training sample be the simplification after examination & approval sample matrix t row data, 1≤
t≤n;
If the global error is more than preset error threshold, the examination & approval model is adjusted, is then back to and executes institute
It states using the examination & approval sample matrix after the simplification to the step examined model and carry out a wheel training, until the overall situation is accidentally
Until difference is less than the error threshold;
If the global error is less than the error threshold, current examination & approval model is determined as the trained examination & approval mould
Type.
6. a kind of computer readable storage medium, the computer-readable recording medium storage has computer-readable instruction, special
Sign is, the examination & approval mould as described in any one of claim 1 to 5 is realized when the computer-readable instruction is executed by processor
The step of type training method.
7. a kind of examination & approval model training terminal device, including memory, processor and it is stored in the memory and can be
The computer-readable instruction run on the processor, which is characterized in that the processor executes the computer-readable instruction
Shi Shixian following steps:
The examination & approval sample of preset number is chosen from history approval record, the examination & approval sample includes that approval results are by just
Sample and approval results are the negative sample of refusal, and the ratio between the number of the positive sample and the number of the negative sample are preset
In ratio range;
Numeralization processing is carried out to information of the examination & approval sample in each examination & approval dimension, the examination & approval sample to be quantized;
By the examination & approval sample composition examination & approval sample matrix of the numeralization, and calculate the covariance square of the examination & approval sample matrix
Battle array, wherein the arbitrary data line of the examination & approval sample matrix is corresponding with the examination & approval sample that one quantizes;
The characteristic value of the covariance matrix of the examination & approval sample matrix is calculated, and chooses the number of preset number from the characteristic value
It is worth maximum characteristic value as dominant eigenvalue;
The examination & approval sample matrix is carried out to simplify processing, the examination & approval sample matrix after being simplified, the examination & approval after the simplification
Only retain row corresponding with the dominant eigenvalue in sample matrix;
Preset examination & approval model is trained using the examination & approval sample matrix after the simplification, obtains trained examination & approval mould
Type.
8. examination & approval model training terminal device according to claim 7, which is characterized in that the examining the numeralization
This composition of lot sample examines sample matrix:
The examination & approval sample of the numeralization is formed into following examination & approval sample matrix:
Wherein, X is the examination & approval sample matrix, xijThe examination & approval sample to quantize for i-th examines the letter in dimension at j-th
Breath, 1≤i≤n, 1≤j≤p, n are the sum of the examination & approval sample of the numeralization, and p is the number of the examination & approval dimension.
9. according to the examination & approval model training terminal device described in claim 8, which is characterized in that described to calculate the examination & approval sample
The covariance matrix of this matrix includes:
The covariance matrix of the examination & approval sample matrix is calculated according to the following formula:
Wherein, R is the covariance matrix of the examination & approval sample matrix, 1≤a≤p, 1≤b≤p.
10. examination & approval model training terminal device according to claim 9, which is characterized in that described to calculate the examination & approval sample
The characteristic value of the covariance matrix of this matrix, and from the characteristic value choose preset number the maximum characteristic value conduct of numerical value
Dominant eigenvalue includes:
Solve characteristic equation | λ I-R |=0, find out eigenvalue λa, wherein I is unit matrix, 1≤a≤p;
The contribution rate of each characteristic value is calculated according to the following formula:
Wherein, ηaIt is characterized value λaContribution rate;
Using the maximum preceding m characteristic value of the numerical value for meeting following condition as dominant eigenvalue:
AndWherein ηthresholdFor preset contribution rate threshold value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810161032.8A CN108446890A (en) | 2018-02-26 | 2018-02-26 | A kind of examination & approval model training method, computer readable storage medium and terminal device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810161032.8A CN108446890A (en) | 2018-02-26 | 2018-02-26 | A kind of examination & approval model training method, computer readable storage medium and terminal device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108446890A true CN108446890A (en) | 2018-08-24 |
Family
ID=63192960
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810161032.8A Pending CN108446890A (en) | 2018-02-26 | 2018-02-26 | A kind of examination & approval model training method, computer readable storage medium and terminal device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108446890A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110390511A (en) * | 2019-06-20 | 2019-10-29 | 深圳壹账通智能科技有限公司 | The credit applications measures and procedures for the examination and approval, device, equipment and storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104616013A (en) * | 2014-04-30 | 2015-05-13 | 北京大学 | Method for acquiring low-dimensional local characteristics descriptor |
CN106204246A (en) * | 2016-08-18 | 2016-12-07 | 易联众信息技术股份有限公司 | A kind of BP neutral net credit estimation method based on PCA |
CN106845330A (en) * | 2016-11-17 | 2017-06-13 | 北京品恩科技股份有限公司 | A kind of training method of the two-dimension human face identification model based on depth convolutional neural networks |
-
2018
- 2018-02-26 CN CN201810161032.8A patent/CN108446890A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104616013A (en) * | 2014-04-30 | 2015-05-13 | 北京大学 | Method for acquiring low-dimensional local characteristics descriptor |
CN106204246A (en) * | 2016-08-18 | 2016-12-07 | 易联众信息技术股份有限公司 | A kind of BP neutral net credit estimation method based on PCA |
CN106845330A (en) * | 2016-11-17 | 2017-06-13 | 北京品恩科技股份有限公司 | A kind of training method of the two-dimension human face identification model based on depth convolutional neural networks |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110390511A (en) * | 2019-06-20 | 2019-10-29 | 深圳壹账通智能科技有限公司 | The credit applications measures and procedures for the examination and approval, device, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10572885B1 (en) | Training method, apparatus for loan fraud detection model and computer device | |
CN108763277B (en) | Data analysis method, computer readable storage medium and terminal device | |
CN111325664B (en) | Style migration method and device, storage medium and electronic equipment | |
CN111368937A (en) | Image classification method and device, and training method, device, equipment and medium thereof | |
CN110264270B (en) | Behavior prediction method, behavior prediction device, behavior prediction equipment and storage medium | |
CN109034206A (en) | Image classification recognition methods, device, electronic equipment and computer-readable medium | |
CN109635939B (en) | Cutting-based convolutional neural network determining method and device | |
CN114332545B (en) | Image data classification method and device based on low-bit pulse neural network | |
CN111709790B (en) | Method, device, equipment and storage medium for identifying abnormal electricity price of day-ahead market | |
WO2021147276A1 (en) | Data processing method and apparatus, and chip, electronic device and storage medium | |
Li et al. | Stock trading strategies based on deep reinforcement learning | |
CN114741507A (en) | Method for establishing and classifying quotation network classification model of graph convolution network based on Transformer | |
CN108446890A (en) | A kind of examination & approval model training method, computer readable storage medium and terminal device | |
CN108765137A (en) | A kind of credit demand prediction technique and system, storage medium | |
CN111353728A (en) | Risk analysis method and system | |
CN111461896A (en) | Method for obtaining underwriting result and related device | |
CN110472246A (en) | Work order classification method, device and storage medium | |
Li et al. | Business email classification using incremental subspace learning | |
KR102409041B1 (en) | portfolio asset allocation reinforcement learning method using actor critic model | |
CN111597927A (en) | Superficial layer convolution neural network-based finger vein recognition system | |
CN111461898A (en) | Method for obtaining underwriting result and related device | |
CN113989566A (en) | Image classification method and device, computer equipment and storage medium | |
CN111476267A (en) | Method and electronic device for classifying drug efficacy according to cell image | |
CN110322150A (en) | A kind of signal auditing method, device and server | |
CN113722440B (en) | Significance analysis method based on keyword recognition and related products |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180824 |