CN109409504A - A kind of data processing method, device, computer and storage medium - Google Patents
A kind of data processing method, device, computer and storage medium Download PDFInfo
- Publication number
- CN109409504A CN109409504A CN201811176811.1A CN201811176811A CN109409504A CN 109409504 A CN109409504 A CN 109409504A CN 201811176811 A CN201811176811 A CN 201811176811A CN 109409504 A CN109409504 A CN 109409504A
- Authority
- CN
- China
- Prior art keywords
- data
- dimensional
- processing
- neural networks
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a kind of data processing method, device, computer and storage mediums.Wherein, mode includes: to obtain one-dimensional initial data, and the one-dimensional initial data is carried out matrixing processing, generates intermediate data, wherein the intermediate data is dimensional matrix data；The intermediate data is input in convolutional neural networks model trained in advance, wherein the convolutional neural networks model includes the convolutional layer of inception structure；The corresponding processing result of the one-dimensional initial data is determined according to the output result of the convolutional neural networks model.The embodiment of the present invention solves the problems, such as that convolutional neural networks model cannot handle one-dimensional data through the above technical solution, further, it include the convolutional layer of inception structure in convolutional neural networks model, feature extraction is carried out to one-dimensional data from various dimensions, more combinations, it improves data characteristics and extracts comprehensive and accuracy, further improve the precision of one-dimensional data processing result.
Description
Technical field
The present embodiments relate to depth learning technology more particularly to a kind of data processing method, device, computer and deposit
Storage media.
Background technique
Convolutional neural networks are a kind of multilayer neural networks, are good at the Machine Learning Problems of processing image class data.Its benefit
Abstract characteristics are extracted from multidimensional data with convolutional layer, then these features are input to classification layer training, obtain suitable model.
Domestic and international image recognition contest champion is obtained using the models for several times of convolutional neural networks algorithm, is computer vision field
Main algorithm, and one of the important algorithm for pushing artificial intelligence to rapidly develop.
Currently, convolutional neural networks are mainly used in field of image processing, since data have local correlations, based on volume
Product neural network can extract the high dimensional feature of image, characteristic pattern be formed, convenient for being analyzed characteristic pattern to obtain result.But
It is that current convolutional neural networks can not handle one-dimensional data.
Summary of the invention
The embodiment of the present invention provides a kind of data processing method, device, computer and storage medium, with realize improve it is one-dimensional
The efficiency of data processing.
In a first aspect, the embodiment of the invention provides a kind of data processing methods, comprising:
One-dimensional initial data is obtained, the one-dimensional initial data is subjected to matrixing processing, generates intermediate data, wherein
The intermediate data is dimensional matrix data；
The intermediate data is input in convolutional neural networks model trained in advance, wherein the convolutional Neural net
Network model includes the convolutional layer of inception structure；
The corresponding processing result of the one-dimensional initial data is determined according to the output result of the convolutional neural networks model.
Second aspect, the embodiment of the invention also provides a kind of data processing equipment, which includes:
One-dimensional data processing module carries out the one-dimensional initial data at matrixing for obtaining one-dimensional initial data
Reason generates intermediate data, wherein the intermediate data is dimensional matrix data；
Intermediate data processing module, for the intermediate data to be input to convolutional neural networks model trained in advance
In, wherein the convolutional neural networks model includes the convolutional layer of inception structure；
Processing result determining module, for determining the one-dimensional original according to the output result of the convolutional neural networks model
The corresponding processing result of beginning data.
The third aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program realizes such as data processing method provided in an embodiment of the present invention when the program is executed by processor.
Fourth aspect, the embodiment of the invention also provides a kind of computers, which is characterized in that including memory, processor
And the computer program that can be run on a memory and in processor is stored, the processor executes real when the computer program
Now such as data processing method provided in an embodiment of the present invention.
One-dimensional initial data by being converted to the intermediate data of two-dimensional matrix format by the embodiment of the present invention, based on preset
Convolutional neural networks model handles intermediate data, obtains the processing result to one-dimensional initial data, solves convolution mind
The problem of one-dimensional data cannot be handled through network model includes further inception structure in convolutional neural networks model
Convolutional layer, from various dimensions, more combinations to one-dimensional data carry out feature extraction, improve data characteristics extract it is comprehensive and
Accuracy further improves the precision of one-dimensional data processing result.
Detailed description of the invention
Fig. 1 is a kind of flow chart for data processing method that the embodiment of the present invention one provides；
Fig. 2 is the schematic diagram of the convolutional layer for the inception structure that the embodiment of the present invention one provides；
Fig. 3 is a kind of flow chart of data processing method provided by Embodiment 2 of the present invention；
Fig. 4 a is a kind of schematic diagram of data sorting mode provided by Embodiment 2 of the present invention；
Fig. 4 b is the schematic diagram of another data sorting mode provided by Embodiment 2 of the present invention；
Fig. 4 c is the schematic diagram of another various data sorting modes provided by Embodiment 2 of the present invention；
Fig. 5 is the flow diagram of the processing method of credit financing data provided by Embodiment 2 of the present invention；
Fig. 6 is a kind of structural schematic diagram for data processing equipment that the embodiment of the present invention three provides；
Fig. 7 is a kind of structural schematic diagram for computer that the embodiment of the present invention four provides.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is a kind of flow chart for data processing method that the embodiment of the present invention one provides, and the present embodiment is applicable to base
Disposition is carried out to one-dimensional data in convolutional neural networks, this method can be filled by data processing provided in an embodiment of the present invention
It sets to execute, specifically comprises the following steps:
Step 110 obtains one-dimensional initial data, and the one-dimensional initial data is carried out matrixing processing, generates mediant
According to, wherein the intermediate data is dimensional matrix data.
The intermediate data is input in convolutional neural networks model trained in advance by step 120, wherein the volume
Product neural network model includes the convolutional layer of inception structure.
Step 130 determines that the one-dimensional initial data is corresponding according to the output result of the convolutional neural networks model
Processing result.
Wherein, the characteristic based on convolutional neural networks model, is generally used for handling image.In the present embodiment,
One-dimensional data is subjected to Data Format Transform, dimensional matrix data, i.e. intermediate data is generated, reaches simulated image data format,
It can be handled based on convolutional neural networks model.Illustratively, one-dimensional initial data include 2500 data, then can be by
One-dimensional initial data is converted to 50 × 50 intermediate data.
Optionally, the one-dimensional initial data is subjected to matrixing processing, generates intermediate data, comprising: according to matrixing
Data sorting in processing rule, determines the sortord of all types of data of one-dimensional initial data；According to the sortord
The one-dimensional initial data is ranked up, generator matrix data, the matrix data is intermediate data.In the present embodiment,
One-dimensional initial data can be including a plurality of types of data, and illustratively, one-dimensional initial data can be but not limited to user
Credit financing data, the credit financing data of user can be including user base data, user behavior data, subscriber's account
Information and user sell letter data etc., since, there are local correlations, matrixing handles number in rule between same type of data
It can be the sortord of data in the sequence of the data type including all types of data and each categorical data according to sequence.
Illustratively, for credit financing data, can be data sorting in matrixing processing rule includes user base data, user
Behavioral data, subscriber's account information and user sell the sorting position of each data in four seed type data of letter data, according to every
The data of above-mentioned four seed type are filled respectively to corresponding sorting position, are generated by the sortord of data in one categorical data
The credit financing data of matrix format, i.e. intermediate data.
In the present embodiment, the intermediate data of matrix format is input to convolutional neural networks model, convolutional neural networks
Include multiple convolutional layers in model, feature extraction, the convolution based on convolutional neural networks model can be carried out to intermediate data respectively
, it can be achieved that extracting the higher dimensional space characteristic relation of intermediate data, realization is determined according to higher dimensional space characteristic relation for the increase of the number of plies
High-precision processing result improves the automation and efficiency of data processing.
Wherein, convolutional neural networks model is that according to great amount of samples, training is completed in advance, such as can be based on supervision
The mode training of study obtains.Sample for training convolutional neural networks model is according to the expectation function of convolutional neural networks model
It can determine that, illustratively, the desired function of convolutional neural networks model is to judge the whether overdue refund of user, then sample information can
Include the credit financing data of user to be, and whether overdue refund.It can be the historical financial credit of acquisition a large number of users
Data and corresponding refund record.Specifically, having the instruction for the convolutional neural networks model for judging the whether overdue refund of user
After the mode of white silk may is that the intermediate data that the historical financial credit data in sample are converted to matrix format, it is input to and does not instruct
In experienced convolutional neural networks model, it can be obtained based on the untrained convolutional neural networks model at the prediction that the model exports
Manage result (whether overdue or overdue probability), when the prediction processing result of model output with refund records different in sample when,
The deviation of record of refunding in the prediction processing result and sample exported based on this model is reversely adjusted in convolutional neural networks model
Weight parameter, iteration executes above-mentioned training process, until the error of convolutional neural networks model is less than default error.It has trained
At convolutional neural networks model have the function of judging the whether overdue refund of user, the current credit financing data of user are defeated
Enter to training complete judge the whether overdue refund of user after, can the high-precision forecast user whether overdue can refund.
It in the present embodiment, include the convolutional layer of at least one layer inception structure in convolutional neural networks model,
The convolutional layer of inception structure can be the sub- convolutional layer including different convolution kernels, and set side by side between each sub- convolutional layer
It sets, i.e., upper one layer of output result is input to each sub- convolutional layer simultaneously, each sub- convolutional layer respectively carries out input data
Convolution, to extract the feature of input data.Optionally, the convolutional layer of the inception structure includes 1 × 1 sub- convolution
Layer, 3 × 3 sub- convolutional layer, empty convolution sum pond layer, wherein described 1 × 1 sub- convolutional layer, described 3 × 3 sub- convolution
Pond layer described in layer, the empty convolution sum is set side by side, and for handling respectively input data, respectively obtains feature square
Battle array.Illustratively, referring to fig. 2, Fig. 2 is the schematic diagram of the convolutional layer of inception structure provided in an embodiment of the present invention.Its
In, the convolution kernel of empty convolution can be 3 × 3 or 5 × 5.Empty convolution can expand the receptive field of convolution, wherein the sense of convolution
The multiple of input are corresponded to by each characteristic value of the wild size for the corresponding input data of each characteristic value, convolutional layer output
Data, illustratively, when convolution kernel is 3 × 3, each characteristic value of each convolutional layer output is corresponded in input data
3 × 3 data.Empty convolution can increase convolution receptive field, when empty convolution convolution kernel be 3 × 3, expansion factor 1, then often
Each characteristic value of one empty convolutional layer output corresponds to 5 × 5 data in input data.It can be expanded by empty convolution
The extraction scope of characteristic relation can extract the characteristic relation between farther away data, avoid since data sorting is to data
The influence of processing result improves data processing precision.
Optionally, the pond layer in the convolutional layer of inception structure can be maximum pond layer, further, maximum
Pond layer can be 3 × 3 maximum pond layer, for carrying out dimension-reduction treatment to input data.
In the present embodiment, process of convolution is carried out to intermediate data by the sub- convolutional layer of different convolution kernels, realized never
Same angle and combination dimension extracts the data characteristics of intermediate data, improves data characteristics and extracts comprehensive and accuracy, into
One step improves the precision of data processed result.
Optionally, input data is handled, after respectively obtaining eigenmatrix, further includes: to the eigenmatrix
Unitized processing is carried out, and eigenmatrix splices by treated, generates the input data of next convolutional layer.Wherein, by
In the sub- convolutional layer of different convolution kernels eigenmatrix it is of different sizes, it is right for the ease of splicing to multiple eigenmatrixes
Characteristic carries out unitized processing, wherein unitized processing can be to be realized by preset algorithm, be can also be and is passed through convolution
Layer is realized.Illustratively, empty convolution sum pond layer is 3 × 3 sub- convolutional layer, then can be after 1 × 1 sub- convolutional layer
3 × 3 pond layer is set, dimensionality reduction is carried out to the eigenmatrix of 1 × 1 sub- convolutional layer.In the present embodiment, to multiple feature squares
Battle array is spliced, and be can be and is spliced according to the sequence of each eigenmatrix to eigenmatrix, and the number in each eigenmatrix
According to sequentially remaining unchanged, the output data of the convolutional layer of inception structure is obtained, the input data as next convolutional layer.
Wherein, next convolutional layer can be the convolutional layer of inception structure.
Optionally, before handling input data, further includes: according to described 1 × 1 sub- convolutional layer, it is described 3 ×
The size of the convolution kernel of pond layer described in 3 sub- convolutional layer, the empty convolution sum respectively expands the input data
Processing, wherein the divergence process includes carrying out boundary zero filling processing to the input data.Illustratively, 50 × 50
Input data carries out boundary zero filling processing respectively, obtains 52 × 52 matrix data, to expand input data.Specifically, can be with
Be to be input to 3 × 3 sub- convolutional layer, 3 × 33 × 3 pond layer of empty convolution sum carries out expansion touch-control, not to input boundary
The input data of zero filling processing carries out divergence process, so that the eigenmatrix size of the output of each sub- convolutional layer is identical.
In some embodiments, it can be before handling input data, to according to demand to input data
Carry out divergence process, and the size of the eigenmatrix based on each sub- convolutional layer output, it is determined whether carry out unitized processing.Work as base
In each sub- convolutional layer output eigenmatrix it is of different sizes when, unitized processing carried out to eigenmatrix, and will treated
Eigenmatrix is spliced, and the input data of next convolutional layer is generated；It is big when the eigenmatrix based on each sub- convolutional layer output
When small identical, directly each eigenmatrix is spliced, generates the input data of next convolutional layer.
Optionally, the convolutional neural networks model further includes full articulamentum, for the feature square according to a upper convolutional layer
Battle array determines processing result.
The technical solution of the present embodiment, by the way that one-dimensional initial data to be converted to the intermediate data of two-dimensional matrix format, base
Intermediate data is handled in preset convolutional neural networks model, obtains the processing result to one-dimensional initial data, is solved
The problem of convolutional neural networks model cannot handle one-dimensional data further, include in convolutional neural networks model
The convolutional layer of inception structure carries out feature extraction to one-dimensional data from various dimensions, more combinations, improves data spy
Sign extracts comprehensive and accuracy, further improves the precision of one-dimensional data processing result.
Embodiment two
In the flow diagram that Fig. 3 is a kind of data processing method provided in an embodiment of the present invention, in above-described embodiment
On the basis of be optimized.Specifically, this method comprises:
S210, one-dimensional initial data is obtained, if handling rule according to matrixing, determines that the one-dimensional initial data exists and lack
Mistake value is then filled processing to the missing values of the one-dimensional initial data.
S220, filled one-dimensional initial data is normalized, the one-dimensional data that obtains that treated.
S230, matrixing processing is carried out to treated the one-dimensional data according at least two data arrays, obtained
To the matrix data of at least two sequences.
S240, each matrix data is determined as to a channel data, to the matrix data of at least two sequence
It is combined, generates the intermediate data at least two channels.
S250, the intermediate data is input in convolutional neural networks model trained in advance, wherein the convolution mind
It include the convolutional layer of inception structure through network model.
S260, the corresponding processing of the one-dimensional initial data is determined according to the output result of the convolutional neural networks model
As a result.
In the present embodiment, matrixing processing rule can be including the size of intermediate data after conversion and all types of
The set-up mode of data, handling rule according to matrixing can determine the normal data amount of all types of data, when the one-dimensional original of acquisition
When all types of data volumes is less than normal data amount in beginning data, determine that there are missing values for one-dimensional initial data.According to all types of
The data class that data include determines the position of missing values.Optionally, zero filling is carried out to missing values, to guarantee the complete of data
Property.Further, filled one-dimensional initial data is normalized, one-dimensional initial data is converted into 0-1 range
Interior data, convenient for the subsequent processing to one-dimensional initial data.
In the present embodiment, according to different arrangement modes, the sequence of different order is carried out to one-dimensional initial data, is obtained
The matrix data of different sortords.Illustratively, one-dimensional initial data can be but not limited to the credit financing data of user,
The credit financing data of user, which can be, sells letter including user base data, user behavior data, subscriber's account information and user
Data.A- Fig. 4 c referring to fig. 4 is the schematic diagram of three kinds of data sorting modes provided in an embodiment of the present invention.Fig. 4 a- Fig. 4 c difference
It is ranked up with credit financing data of the different sortords to user, obtains the matrix data of three kinds of different sequences.Wherein,
It include multiple data in the data of each type, illustratively, user base data can be including but not limited to user
The basic informations such as gender, account；When user behavior data can be the including but not limited to number of logging in system by user, login
Between, the information such as action type after login system；When subscriber's account information can be including but not limited to customer consumption information, consumption
Between etc. information；User, which sells letter data, can be the including but not limited to information such as credit line, debt-credit time.For each type
Data, can be according to preset sequence be arranged in corresponding region, can also be and be randomly dispersed in corresponding region.It is exemplary
, letter data is sold in Fig. 4 a- Fig. 4 c in the sortord difference of the data of each of Fig. 4 a- Fig. 4 c type, such as user
In sequence it is different.It should be noted that the three kinds of sortords provided in Fig. 4 a- Fig. 4 c are not limited in the present embodiment, at it
It can also be other sortords in his embodiment.
After generating the matrix data of multiple sortords, the matrix data of multiple sortords is combined, it is raw
At intermediate data.Specifically, being combined to multiple matrix datas, shape using each matrix data as a channel data
The intermediate data of formula multichannel, and the matrix data of multichannel is input in convolutional neural networks.Wherein, convolutional neural networks
Each convolutional layer of model carries out feature extraction to intermediate data, and the eigenmatrix of multichannel, convolutional neural networks model can be obtained
Full articulamentum data processed result is determined according to the matrix character of multichannel, such as can be to the matrix data of multichannel into
After row mean value or weighted mean processing, data processed result is further obtained.
The technical solution of the present embodiment, the matrixing by carrying out different sequences to one-dimensional initial data are handled, are obtained more
The intermediate data in channel, avoid the problem that the single obtained intermediate data of sorting since sortord influences data processing precision,
Convenient for the characteristic relation of convolutional neural networks model extraction different spaces sequence, the comprehensive of characteristic relation extraction is improved.
The data processing method of the present embodiment, which can be, to be applied and financial air control field.In financial air control field, usually
Using modes such as scorecard, logistic regression, decision trees, but against the increase of data dimension, above-mentioned conventional method is handled
When, model becomes increasingly complex, time-consuming increase, and combine high dimensional feature it is also more and more difficult, can not quickly and effectively to data into
Row processing.Data processing method based on the embodiment of the present invention handles finance data, and not needing any code can
Machine learning model and model report are obtained, and convolutional neural networks model can extract higher dimensional space characteristic to finance data
According to combination more multivariable improves the automation and high efficiency of modeling and data processing.
Fig. 5 is the flow diagram of the processing method of credit financing data provided by Embodiment 2 of the present invention, in a reality
It applies in example, carrying out data processing to credit financing data includes:
S310, one-dimensional credit financing data are obtained, if handling rule according to matrixing, determines the credit financing data
There are missing values, then are filled processing to the missing values of the credit financing data, credit financing data can be including with
Family basic data, user behavior data, subscriber's account information and user sell letter data.
S320, filled credit financing data are normalized, the credit financing data that obtain that treated.
S330, treated the credit financing data are carried out at matrixing according at least two data arrays
Reason, obtains the matrix data of at least two sequences.
S340, each matrix data is determined as to a channel data, to the matrix data of at least two sequence
It is combined, generates the intermediate data at least two channels.
S350, the intermediate data is input in convolutional neural networks model trained in advance, wherein the convolution mind
Has the function of the overdue refund probability for judging user through network model, the convolutional neural networks model includes inception
The convolutional layer of structure, convolutional neural networks model have the overdue refund probability for judging user.
S360, determine that the credit financing data are corresponding overdue according to the output result of the convolutional neural networks model
Refund probability.
The technical solution of the present embodiment is handled finance data by way of establishing convolutional neural networks model,
The stability and good ageing of convolutional neural networks model, and higher dimensional space feature can be extracted to the finance data of matrixing and closed
System, solves that conventional model modeling is time-consuming, and the problem of can not extracting high dimensional feature has auto-modeling, high efficiency and place
Manage result advantage with high accuracy.
Embodiment three
Fig. 6 is a kind of structural schematic diagram for data processing equipment that the embodiment of the present invention three provides, the data processing equipment
Data processing method is provided for executing any embodiment of that present invention, comprising:
The one-dimensional initial data is carried out matrixing for obtaining one-dimensional initial data by one-dimensional data processing module 410
Processing generates intermediate data, wherein the intermediate data is dimensional matrix data；
Intermediate data processing module 420, for the intermediate data to be input to convolutional neural networks mould trained in advance
In type, wherein the convolutional neural networks model includes the convolutional layer of inception structure；
Processing result determining module 430, for determining described one according to the output result of the convolutional neural networks model
Tie up the corresponding processing result of initial data.
Optionally, the one-dimensional initial data is the credit financing data of user, the corresponding place of the one-dimensional initial data
Manage the overdue probability of refund that result is the user.
Optionally, one-dimensional data processing module 410 includes:
Sortord determination unit determines the one-dimensional original number for handling data sorting in rule according to matrixing
According to the sortord of all types of data；
Intermediate data determination unit is generated for being ranked up the one-dimensional initial data according to the sortord
Matrix data, the matrix data are intermediate data.
Optionally, the data sorting is at least two, obtains the matrix data of at least two sequences；Correspondingly, one-dimensional
Data processing module 410 further include:
Data combined segment member, for after generator matrix data, each matrix data to be determined as a port number
According to the matrix data to sort to described at least two is combined, and generates the intermediate data at least two channels.
Optionally, described device further include:
Database population module, for carrying out matrix to the one-dimensional initial data according at least two data arrays
Before change processing, if determining that there are missing values for the one-dimensional initial data, then to described one-dimensional original according to preset data form
The missing values of data are filled processing；
Normalized module obtains that treated for filled one-dimensional initial data to be normalized
One-dimensional data；
Correspondingly, one-dimensional data processing module 410 is used for, according at least two data arrays, treated to described
One-dimensional data carries out matrixing processing.
Optionally, the convolutional layer of the inception structure includes 1 × 1 sub- convolutional layer, 3 × 3 sub- convolutional layer, sky
Hole convolution sum pond layer, wherein described 1 × 1 sub- convolutional layer, described 3 × 3 sub- convolutional layer, described in the empty convolution sum
Pond layer is set side by side, and for handling respectively input data, respectively obtains eigenmatrix.
Optionally, the convolutional layer of the inception structure further include:
Eigenmatrix splicing module, for handling input data, after respectively obtaining eigenmatrix, to the spy
Sign matrix carries out unitized processing, and by treated, eigenmatrix splices, and generates the input data of next convolutional layer.
Optionally, the convolutional layer of the inception structure further include: further include:
Data augmentation module, for before handling input data, according to described 1 × 1 sub- convolutional layer, described
The size of the convolution kernel of pond layer described in 3 × 3 sub- convolutional layer, the empty convolution sum respectively expands the input data
Open processing, wherein the divergence process includes carrying out boundary zero filling processing to the input data.
Optionally, the convolutional neural networks model further includes full articulamentum, for the feature square according to a upper convolutional layer
Battle array determines processing result.
Data processing provided by any embodiment of the invention can be performed in data processing equipment provided in an embodiment of the present invention
Method has the corresponding functional module of configuration for executing data processing and beneficial effect.
Example IV
Fig. 7 is a kind of structural schematic diagram for computer that the embodiment of the present invention four provides.Fig. 7, which is shown, to be suitable for being used to realizing
The block diagram of the illustrative computer 12 of embodiment of the present invention.The computer 12 that Fig. 7 is shown is only an example, should not be to this
The function and use scope of inventive embodiments bring any restrictions.
As shown in fig. 7, computer 12 is showed in the form of universal computing device.The component of computer 12 may include but not
Be limited to: one or more processor or processing unit 16, system storage 28 connect different system components (including system
Memory 28 and processing unit 16) bus 18.
Bus 18 indicates one of a few class bus structures or a variety of, including memory bus or Memory Controller,
Peripheral bus, graphics acceleration port, processor or the local bus using any bus structures in a variety of bus structures.It lifts
For example, these architectures include but is not limited to industry standard architecture (ISA) bus, microchannel architecture (MAC)
Bus, enhanced isa bus, Video Electronics Standards Association (VESA) local bus and peripheral component interconnection (PCI) bus.
Computer 12 typically comprises a variety of computer system readable media.These media can be and any can be calculated
The usable medium that machine 12 accesses, including volatile and non-volatile media, moveable and immovable medium.
System storage 28 may include the computer system readable media of form of volatile memory, such as arbitrary access
Memory (RAM) 30 and/or cache memory 32.Computer 12 may further include other removable/nonremovable
, volatile/non-volatile computer system storage medium.Only as an example, storage system 34 can be used for reading and writing not removable
Dynamic, non-volatile magnetic media (Fig. 7 do not show, commonly referred to as " hard disk drive ").Although being not shown in Fig. 7, can provide
Disc driver for being read and write to removable non-volatile magnetic disk (such as " floppy disk "), and to moving non-volatile light
The CD drive of disk (such as CD-ROM, DVD-ROM or other optical mediums) read-write.In these cases, each driver
It can be connected by one or more data media interfaces with bus 18.Memory 28 may include that at least one program produces
Product, the program product have one group of (for example, at least one) program module, these program modules are configured to perform of the invention each
The function of embodiment.
Program/utility 40 with one group of (at least one) program module 42 can store in such as memory 28
In, such program module 42 include but is not limited to operating system, one or more application program, other program modules and
It may include the realization of network environment in program data, each of these examples or certain combination.Program module 42 is usual
Execute the function and/or method in embodiment described in the invention.
Computer 12 can also be logical with one or more external equipments 14 (such as keyboard, sensing equipment, display 24 etc.)
Letter, can also be enabled a user to one or more equipment interact with the computer 12 communicate, and/or with make the computer
The 12 any equipment (such as network interface card, modem etc.) that can be communicated with one or more of the other calculating equipment communicate.
This communication can be carried out by input/output (I/O) interface 22.Also, computer 12 can also pass through network adapter 20
With one or more network (such as local area network (LAN), wide area network (WAN) and/or public network, such as internet) communication.
As shown, network adapter 20 is communicated by bus 18 with other modules of computer 12.It should be understood that although not showing in figure
Out, can in conjunction with computer 12 use other hardware and/or software module, including but not limited to: microcode, device driver,
Redundant processing unit, external disk drive array, RAID system, tape drive and data backup storage system etc..
Processing unit 16 by the program that is stored in system storage 28 of operation, thereby executing various function application and
Data processing, such as realize following method:
One-dimensional initial data is obtained, the one-dimensional initial data is subjected to matrixing processing, generates intermediate data, wherein
The intermediate data is dimensional matrix data；
The intermediate data is input in convolutional neural networks model trained in advance, wherein the convolutional Neural net
Network model includes the convolutional layer of inception structure；
The corresponding processing result of the one-dimensional initial data is determined according to the output result of the convolutional neural networks model.
Embodiment five
The embodiment of the present invention five additionally provides a kind of computer readable storage medium.It is stored thereon with computer program, it should
Program is executed by processor following method:
One-dimensional initial data is obtained, the one-dimensional initial data is subjected to matrixing processing, generates intermediate data, wherein
The intermediate data is dimensional matrix data；
The intermediate data is input in convolutional neural networks model trained in advance, wherein the convolutional Neural net
Network model includes the convolutional layer of inception structure；
The corresponding processing result of the one-dimensional initial data is determined according to the output result of the convolutional neural networks model.
The computer storage medium of the embodiment of the present invention, can be using any of one or more computer-readable media
Combination.Computer-readable medium can be computer-readable signal media or computer readable storage medium.It is computer-readable
Storage medium for example may be-but not limited to-the system of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, device or
Device, or any above combination.The more specific example (non exhaustive list) of computer readable storage medium includes: tool
There are electrical connection, the portable computer diskette, hard disk, random access memory (RAM), read-only memory of one or more conducting wires
(ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.In this document, computer-readable storage
Medium can be any tangible medium for including or store program, which can be commanded execution system, device or device
Using or it is in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
Further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on a remote computer or server completely on the remote computer on the user computer.?
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or
Wide area network (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as mentioned using Internet service
It is connected for quotient by internet).
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (12)
1. a kind of data processing method characterized by comprising
One-dimensional initial data is obtained, the one-dimensional initial data is subjected to matrixing processing, generates intermediate data, wherein described
Intermediate data is dimensional matrix data；
The intermediate data is input in convolutional neural networks model trained in advance, wherein the convolutional neural networks mould
Type includes the convolutional layer of inception structure；
The corresponding processing result of the one-dimensional initial data is determined according to the output result of the convolutional neural networks model.
2. the method according to claim 1, wherein the one-dimensional initial data is the credit financing number of user
According to the corresponding processing result of the one-dimensional initial data is the overdue probability of refund of the user.
3. the method according to claim 1, wherein the one-dimensional initial data is carried out matrixing processing, life
At intermediate data, comprising:
Data sorting in rule is handled according to matrixing, determines the sortord of all types of data of one-dimensional initial data；
The one-dimensional initial data is ranked up according to the sortord, generator matrix data, during the matrix data is
Between data.
4. according to the method described in claim 3, it is characterized in that, the data sorting be at least two, obtain at least two
The matrix data of sequence；Correspondingly, after generator matrix data, further includes:
Each matrix data is determined as a channel data, the matrix data of at least two sequence is combined,
Generate the intermediate data at least two channels.
5. according to the method described in claim 3, it is characterized in that, according at least two data arrays to described one-dimensional
Initial data carries out before matrixing processing, further includes:
If determining that there are missing values for the one-dimensional initial data, then to the one-dimensional initial data according to preset data form
Missing values are filled processing；
Filled one-dimensional initial data is normalized, the one-dimensional data that obtains that treated；
Correspondingly, carrying out matrixing processing to the one-dimensional initial data according at least two data arrays, comprising:
Matrixing processing is carried out to treated the one-dimensional data according at least two data arrays.
6. -5 any method according to claim 1, which is characterized in that the convolutional layer of the inception structure includes 1
× 1 sub- convolutional layer, 3 × 3 sub- convolutional layer, empty convolution sum pond layer, wherein described 1 × 1 sub- convolutional layer, described 3 ×
Pond layer described in 3 sub- convolutional layer, the empty convolution sum is set side by side, for handling respectively input data, respectively
Obtain eigenmatrix.
7. according to the method described in claim 6, respectively obtaining eigenmatrix it is characterized in that, handle input data
Later, further includes:
Unitized processing is carried out to the eigenmatrix, and eigenmatrix splices by treated, generates next convolutional layer
Input data.
8. according to the method described in claim 6, it is characterized in that, before handling input data, further includes:
According to the convolution of pond layer described in described 1 × 1 sub- convolutional layer, described 3 × 3 sub- convolutional layer, the empty convolution sum
The size of core respectively to the input data carry out divergence process, wherein the divergence process include to the input data into
Row bound zero filling processing.
9. -5 any method according to claim 1, which is characterized in that the convolutional neural networks model further includes connecting entirely
Layer is connect, for determining processing result according to the eigenmatrix of a upper convolutional layer.
10. a kind of data processing equipment characterized by comprising
The one-dimensional initial data is carried out matrixing processing for obtaining one-dimensional initial data by one-dimensional data processing module, raw
At intermediate data, wherein the intermediate data is dimensional matrix data；
Intermediate data processing module, for the intermediate data to be input in convolutional neural networks model trained in advance,
In, the convolutional neural networks model includes the convolutional layer of inception structure；
Processing result determining module, for determining the one-dimensional original number according to the output result of the convolutional neural networks model
According to corresponding processing result.
11. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The data processing method as described in any in claim 1-9 is realized when execution.
12. a kind of computer, which is characterized in that including memory, processor and storage can be transported on a memory and in processor
Capable computer program, the processor realize the data as described in claim 1-9 is any when executing the computer program
Processing method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811176811.1A CN109409504A (en) | 2018-10-10 | 2018-10-10 | A kind of data processing method, device, computer and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811176811.1A CN109409504A (en) | 2018-10-10 | 2018-10-10 | A kind of data processing method, device, computer and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109409504A true CN109409504A (en) | 2019-03-01 |
Family
ID=65467364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811176811.1A Pending CN109409504A (en) | 2018-10-10 | 2018-10-10 | A kind of data processing method, device, computer and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109409504A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110263916A (en) * | 2019-05-31 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Data processing method and device, storage medium and electronic device |
| CN110599452A (en) * | 2019-08-07 | 2019-12-20 | 全球能源互联网研究院有限公司 | Rust detection network, method, computer device and readable storage medium |
| CN111386697A (en) * | 2019-04-02 | 2020-07-07 | 深圳市大疆创新科技有限公司 | Overexposure recovery processing method, overexposure recovery processing device and computer-readable storage medium |
| CN111652627A (en) * | 2020-07-07 | 2020-09-11 | 中国银行股份有限公司 | Risk assessment method and device |
| CN112669142A (en) * | 2021-01-05 | 2021-04-16 | 深圳前海微众银行股份有限公司 | Modeling method, device and equipment of high-dimensional behavior data and readable storage medium |
| CN112801712A (en) * | 2021-02-09 | 2021-05-14 | 杭州网易再顾科技有限公司 | Advertisement putting strategy optimization method and device |
| CN113919405A (en) * | 2020-07-07 | 2022-01-11 | 华为技术有限公司 | Data processing method and device and related equipment |
-
2018
- 2018-10-10 CN CN201811176811.1A patent/CN109409504A/en active Pending
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111386697A (en) * | 2019-04-02 | 2020-07-07 | 深圳市大疆创新科技有限公司 | Overexposure recovery processing method, overexposure recovery processing device and computer-readable storage medium |
| CN110263916A (en) * | 2019-05-31 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Data processing method and device, storage medium and electronic device |
| CN110599452A (en) * | 2019-08-07 | 2019-12-20 | 全球能源互联网研究院有限公司 | Rust detection network, method, computer device and readable storage medium |
| CN110599452B (en) * | 2019-08-07 | 2022-02-22 | 全球能源互联网研究院有限公司 | Rust detection system, method, computer device and readable storage medium |
| CN111652627A (en) * | 2020-07-07 | 2020-09-11 | 中国银行股份有限公司 | Risk assessment method and device |
| CN113919405A (en) * | 2020-07-07 | 2022-01-11 | 华为技术有限公司 | Data processing method and device and related equipment |
| CN113919405B (en) * | 2020-07-07 | 2024-01-19 | 华为技术有限公司 | Data processing method and device and related equipment |
| CN111652627B (en) * | 2020-07-07 | 2024-04-23 | 中国银行股份有限公司 | Risk assessment method and risk assessment device |
| CN112669142A (en) * | 2021-01-05 | 2021-04-16 | 深圳前海微众银行股份有限公司 | Modeling method, device and equipment of high-dimensional behavior data and readable storage medium |
| CN112669142B (en) * | 2021-01-05 | 2024-02-20 | 深圳前海微众银行股份有限公司 | Modeling method, device and equipment for high-dimensional behavior data and readable storage medium |
| CN112801712A (en) * | 2021-02-09 | 2021-05-14 | 杭州网易再顾科技有限公司 | Advertisement putting strategy optimization method and device |
| CN112801712B (en) * | 2021-02-09 | 2024-03-01 | 杭州网易再顾科技有限公司 | Advertisement putting strategy optimization method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109409504A (en) | A kind of data processing method, device, computer and storage medium | |
| CN112633419B (en) | Small sample learning method and device, electronic equipment and storage medium | |
| CN109522942B (en) | Image classification method and device, terminal equipment and storage medium | |
| CN108959246A (en) | Answer selection method, device and electronic equipment based on improved attention mechanism | |
| CN113822209B (en) | Hyperspectral image recognition method and device, electronic equipment and readable storage medium | |
| CN107832432A (en) | A kind of search result ordering method, device, server and storage medium | |
| CN114676704B (en) | Sentence emotion analysis method, device and equipment and storage medium | |
| CN111694937A (en) | Interviewing method and device based on artificial intelligence, computer equipment and storage medium | |
| CN109766557A (en) | A kind of sentiment analysis method, apparatus, storage medium and terminal device | |
| CN108509407A (en) | Text semantic similarity calculating method, device and user terminal | |
| CN111027576A (en) | Cooperative significance detection method based on cooperative significance generation type countermeasure network | |
| WO2023116111A1 (en) | Disk fault prediction method and apparatus | |
| CN111639690A (en) | Fraud analysis method, system, medium, and apparatus based on relational graph learning | |
| CN111782804B (en) | Text CNN-based co-distributed text data selection method, system and storage medium | |
| CN115344805A (en) | Material auditing method, computing equipment and storage medium | |
| KR102401115B1 (en) | Artificial neural network Automatic design generation apparatus and method using UX-bit, skip connection structure and channel-wise concatenation structure | |
| CN110197213B (en) | Image matching method, device and equipment based on neural network | |
| CN115266141B (en) | Method and device for detecting spot welding quality based on GRU-C network and storage medium | |
| CN113010687B (en) | Exercise label prediction method and device, storage medium and computer equipment | |
| CN114741697A (en) | Malicious code classification method and device, electronic equipment and medium | |
| CN111382246B (en) | Text matching method, matching device, terminal and computer readable storage medium | |
| CN116303559B (en) | Method, system and storage medium for controlling form question and answer | |
| CN117368705A (en) | Digital integrated circuit time sequence monitoring method based on graph convolution neural network | |
| CN116701619A (en) | Form information extraction method, system and equipment based on compressed pre-training language model | |
| CN117911785A (en) | Heterogeneous multi-mode image classification method and device based on missing mode |
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 |
Application publication date: 20190301 |
|
| RJ01 | Rejection of invention patent application after publication |