CN109086837A - User property classification method, storage medium, device and electronic equipment based on convolutional neural networks - Google Patents
User property classification method, storage medium, device and electronic equipment based on convolutional neural networks Download PDFInfo
- Publication number
- CN109086837A CN109086837A CN201811246303.6A CN201811246303A CN109086837A CN 109086837 A CN109086837 A CN 109086837A CN 201811246303 A CN201811246303 A CN 201811246303A CN 109086837 A CN109086837 A CN 109086837A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- user property
- user
- measurand
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of user property classification method, storage medium, device and electronic equipment based on convolutional neural networks.User property classification method of the present invention based on convolutional neural networks includes the following steps: using the psychological attribute related electronic data of multiple users as input vector collection input user property classification convolutional neural networks model；Using the Psychological Evaluation result of the multiple user obtained by questionnaire as object vector collection input user property classification convolutional neural networks model, and update user property classification convolutional neural networks model；The psychological attribute related electronic data of measurand is inputted into user property classification convolutional neural networks model, and obtains the attributive classification result of measurand from the output of user property classification convolutional neural networks model.User property classification method of the present invention based on convolutional neural networks is easy to operate, can improve the efficiency and accuracy of user property classification.
Description
Technical field
The present invention relates to user property classification fields, more particularly to the user property classification side based on convolutional neural networks
Method, storage medium, device and electronic equipment.
Background technique
At present internet be newly sold, precisely popularization, recruitment, reference, love and marriage matching, occupational planning, individualized education, product
In the intelligent Applications such as customization, it usually needs accurately obtain personal attribute's feature of user, and make user by doing questionnaire
Mode obtains personal attribute's feature of user, and by shifting to an earlier date printed table, or online webpage provides Research advancement on measuring scale questionnaire
(list of scale and option) is calculated specific by submitting option to provide assessment result according to the relationship of index model and scale
Each dimension appraisal result of personality model.This method requires user to read all scales and option, makes judge one by one and selects
It selects, usual time-consuming 10-30 minutes, just needs to fill in more set questionnaires when needing with multiple personality models, takes more time,
And appraiser is also required to take considerable time statistics scoring, Research advancement on measuring scale is also inconvenient for same measurand in a short time
In reuse.And the participation enthusiasm of the user of questionnaire survey is not high, and several answers is often at will selected to complete questionnaire, according to this
The questionnaire that sample is answered, can not accurately know that personal attribute's feature of user, the accuracy of questionnaire not can guarantee, can not be accurate
Reach target.
Convolutional neural networks are one of the network structures of great representative in depth learning technology, are obtained in data processing field
Very big success, on the ImageNet data set of international standard, many successful models are all based on CNN
(Constitutional Neural Networks, convolutional neural networks).CNN is compared to traditional image processing algorithm
One of advantage is, avoids preprocessing process early period (extracting manual features etc.) to image complexity, can directly input original
Beginning data.
Summary of the invention
Based on this, the user property classification method based on convolutional neural networks that the object of the present invention is to provide a kind of, this
The user property classification method based on convolutional neural networks of invention is easy to operate, can improve the efficiency and standard of user property classification
True property.
The present invention is implemented by following scheme:
A kind of user property classification method based on convolutional neural networks, includes the following steps:
Using the psychological attribute related electronic data of multiple users as input vector collection input user property classification convolution mind
Through network model；
User property is inputted using the Psychological Evaluation result of the multiple user obtained by questionnaire as object vector collection
Classification convolutional neural networks model, and update user property classification convolutional neural networks model；
By the psychological attribute related electronic data input user property classification convolutional neural networks model of measurand, and from
The attributive classification result of user property classification convolutional neural networks model output measurand.
User property classification method based on convolutional neural networks of the invention, it is easy to operate, it can largely save tested pair
As the time with appraiser, and convenient for being reused in the short time to same measurand, moreover it is possible to utilize NPU, GPU, TPU etc.
The higher-dimension computation capability of hardware.By using the psychological attribute related electronic data of user as input vector collection, by questionnaire
The Psychological Evaluation result of acquisition, to training user's attributive classification convolutional neural networks model, is used as the input of object vector collection
The attributive classification result that family attributive classification convolutional neural networks model obtains is more accurate, improves the efficiency of user property classification
And accuracy, more accurate user attribute data can be provided for the intelligent Application of internet.
Further, the psychological attribute related electronic data includes the facial picture of user and/or the person's handwriting figure of user
The video image of the audio and/or user's walking of picture and/or user's speech.
Further, it is inputted the Psychological Evaluation result of the multiple user obtained by questionnaire as object vector collection
User property is classified before convolutional neural networks model, is further included the steps that the qualified questionnaire evaluating result of screening, is specifically included:
The first deviation in questionnaire between the answer of same problems is obtained, and/or obtains the adjacent problem of sequence in questionnaire
Answer between the second deviation；
If first deviation is greater than the second setting threshold less than the first given threshold and/or second deviation
Value then determines that the evaluating result of the same problems and/or the adjacent problem of the sequence is qualified.
Will the qualified questionnaire evaluating result of screening as the target of training user's attributive classification convolutional neural networks model to
Quantity set, the result that family attributive classification convolutional neural networks model can be used are more accurate.
Further, if the category of the measurand obtained from the output of user property classification convolutional neural networks model
Property classification results and pass through questionnaire obtain measurand Psychological Evaluation result between difference be more than third given threshold, then
The Psychological Evaluation result for the measurand that the questionnaire is obtained is inputted as object vector, and updates user property classification convolution mind
Through network model.
By being used to update user property classification convolution mind for the big psychological attribute related electronic data of evaluating result deviation
Through network model, the assessment that family attributive classification convolutional neural networks model can be used is more accurate.
Further, the Psychological Evaluation result of the measurand which obtained is inputted as object vector, and is updated
User property classification convolutional neural networks model, specifically comprises the following steps:
By the psychological attribute related electronic data of the measurand and from user property classification convolutional neural networks model it is defeated
The attributive classification result of measurand out is put into alternative training set；
Obtain the attributive classification knot of the measurand obtained from the output of user property classification convolutional neural networks model
Difference between the Psychological Evaluation result for the measurand that fruit and questionnaire obtain；
By the difference be more than third given threshold the measurand psychological attribute related electronic data and by asking
The Psychological Evaluation result for rolling up the measurand obtained is put into repetitive exercise collection；
User property classification convolutional neural networks model is updated according to repetitive exercise collection.
Further, the attributive classification result of the measurand includes all dimensions for trained Psychological Evaluation result
Degree, and can be according to the Psychological Evaluation type change for training.
Further, the optional range of the answer of the questionnaire is continuous real-value range.
Further, the present invention also provides a kind of computer can storage medium, store computer program thereon, the calculating
The user property classification side based on convolutional neural networks as described in above-mentioned any one is realized when machine program is executed by processor
The step of method.
Further, the present invention also provides a kind of user property sorters, comprising:
Input module, for belonging to the psychological attribute related electronic data of multiple users as input vector collection input user
Property classification convolutional neural networks model；
Feedback training module, the Psychological Evaluation result of the multiple user for that will be obtained by questionnaire as target to
Quantity set inputs training feedback model, and updates user property classification convolutional neural networks model；
User property classification convolutional neural networks module, for inputting the psychological attribute related electronic data of measurand
User property classify convolutional neural networks model, and from user property classification convolutional neural networks model output measurand category
Property classification results.
Further, it the present invention also provides a kind of electronic equipment, including controller and memory, is stored on the memory
Have computer program, when which is executed by processor realize as described in above-mentioned any one based on convolutional Neural net
The step of user property classification method of network.
In order to better understand and implement, the invention will now be described in detail with reference to the accompanying drawings.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the user property classification method based on convolutional neural networks in embodiment；
Fig. 2 is convolutional neural networks schematic diagram in a kind of embodiment；
Fig. 3 is questionnaire topic schematic diagram in a kind of embodiment；
Fig. 4 is the flow chart that convolutional neural networks model is updated in a kind of embodiment；
Fig. 5 is user property sorter overall structure diagram in a kind of embodiment；
Fig. 6 is user property sorter training structure schematic diagram in a kind of embodiment；
Fig. 7 is user property sorter schematic diagram of application structure in a kind of embodiment；
Fig. 8 is electronic devices structure schematic diagram in a kind of embodiment.
Specific embodiment
Referring to Fig. 1, it is the user property classification method based on convolutional neural networks in an embodiment of the present invention
Flow chart.The user property classification method based on convolutional neural networks, includes the following steps:
Step S10: using the psychological attribute related electronic data of multiple users as input vector collection input user property point
Class convolutional neural networks model.
User property classification, to a series of assessment that people carries out, is obtained pair according to certain theories of psychology
The psychology such as ability, personality and mental health of people and a series of quantifications of behavioral trait are measured, and are surveyed comprising aptitude tests, personality
Examination and interest test etc., in the present embodiment, the user property classification mainly include that personality is tested.
The measurand is currently to need to carry out the user of user property classification.
The electronic data includes by electronically, and optical instrument or similar means generate, the information of storage or transmitting,
In the present embodiment, the electronic data is the data information of electronic form.The psychological attribute related electronic data of the user is
Data information relevant to the user for carrying out user property classification is currently needed is formed by electronic data.
The convolutional neural networks are the depth machine learning method based on artificial neural network, by special training
Afterwards, the electronic data information of input can be identified.In the present embodiment, the user property classification convolutional neural networks model is
After trained, user property classification results can be obtained according to the user property related electronic data inputted.
Step S11: it is inputted the Psychological Evaluation result of the multiple user obtained by questionnaire as object vector collection
User property classification convolutional neural networks model, and update user property classification convolutional neural networks model.
In the present embodiment, the user property classification results are the characteristic amount of progress to psychological aspects such as the personality of user
Change evaluation as a result, update user property classification convolutional neural networks model pass through input vector collection and object vector collection to
Family attributive classification convolutional neural networks are trained, the user property classification convolutional neural networks updated after training.
Step S12: by the psychological attribute related electronic data input user property classification convolutional neural networks of measurand
Model, and from user property classification convolutional neural networks model output measurand attributive classification result.
For the accuracy for guaranteeing test result, the questionnaire in the present embodiment refers to comparison comprehensively and the questionnaire of specification, further
, the questionnaire in the present embodiment further includes the more fully scale with specification.
Wherein, input vector integrates as the input of user property classification convolutional neural networks model, including voice, text, figure
Picture and video information etc., and the object vector collection of user property classification convolutional neural networks model is the reason of current input vector collection
Want to export.In the present embodiment, need to collect a large amount of psychological attribute related electronic data, and pass through the corresponding of questionnaire acquisition
Psychological Evaluation is as a result, and be put into training set for the psychological attribute related electronic data of above-mentioned user and corresponding Psychological Evaluation result
In, to be trained to user property classification convolutional neural networks model.By different input vector collection and its is corresponding
Object vector collection is trained, and is constantly updated the coefficient matrix of the user property classification convolutional neural networks model, is made it
Attributive classification result is more and more accurate.
User property classification method based on convolutional neural networks of the invention, it is easy to operate, it can largely save tested pair
As the time with appraiser, and convenient for being reused in the short time to same measurand, moreover it is possible to utilize NPU, GPU, TPU etc.
The higher-dimension computation capability of hardware.By using the psychological attribute related electronic data of user as input vector collection, by questionnaire
The Psychological Evaluation result of acquisition, to training user's attributive classification convolutional neural networks model, is used as the input of object vector collection
The attributive classification result that family attributive classification convolutional neural networks model obtains is more accurate, improves the efficiency of user property classification
And accuracy, more accurate user attribute data can be provided for the intelligent Application of internet.
It is all customized that convolutional neural networks model of the present invention can be each layer, can also be at some existing points
Add folded form on the basis of class model.Referring to Fig. 2, in one embodiment, using inception network as user property
Classify convolutional neural networks model, it is trained after inception network, input the psychological attribute related electronic data of user
Afterwards, exportable user property classification results.
In one embodiment, the psychological attribute related electronic data includes facial picture and/or the user of user
The video image of the audio and/or user's walking of handwriting image and/or user's speech.
Wherein, the facial picture embodies the facial characteristics of user；The handwriting image embodies the person's handwriting row of user
Away from, the shape of inclined direction, font size, shape, speed, dynamics, layout and stroke and lines；The audio embodies user
Fundamental frequency, harmonic to noise ratio, the formant of the voice of speech；The video image of the walking embodies user's body characteristic point, width
The features such as height ratio, profile length and enclosed area ratio, eccentricity.User property classification convolutional neural networks model energy according to
Features described above analyzes user, obtains the user property classification results of user.
In one embodiment, the user property classification results include the extropism of user, doing one's duty property, emotionality, open
Putting property and pleasant property.Due to the universality of method, the user property inspection result can also refer to comprising more psychology or personality
Mark.Such as " flare-introversion " of MBTI, " feeling-intuition ", " thinking-emotion ", " judgement-consciousness "；The domination of DISC
Property, influence power, compliance, stability；Big seven positive emotionality, negative valency, nominal price, negative-morality, reliability, agreeable, following property；
A, B, C, E, F, G, H, I, L, M, N, O, Q1, Q2, Q3, Q4 of 16PF；Hs, D, Hy, Pd, Mf, Pa, Pt, Sc, Ma, Si of MMPI；
The inside and outside tropism of EPQ, nervousness, psychoticism, the property covered up.
User property classification method based on convolutional neural networks of the invention, can pass through the electronic data of a user
The above-mentioned testing result of any one is detected, multinomial above-mentioned detection knot can also be detected by the electronic data of a user
Fruit.
In the above-described embodiments, if to user property classification convolutional neural networks model training used by questionnaire obtain
The Psychological Evaluation result inaccuracy taken, then will affect trained effect, and further influence the correct of user property classification results
Property.Therefore, in one embodiment, step S11 is being carried out, i.e., using the user psychology evaluating result obtained by questionnaire as mesh
Before marking vector set input user property classification convolutional neural networks model, also included the following steps:
Step S111: qualified questionnaire evaluating result is screened.
The questionnaire evaluating result of the qualification refers to the evaluating result that can accurately react the psychological condition of user.
In the present embodiment, qualified questionnaire evaluating result is screened, is specifically comprised the following steps:
Step S112: obtaining the first deviation in questionnaire between the answer of same problems, and/or obtains sequence in questionnaire
The second deviation between the answer of adjacent problem.
Step S113: if first deviation is greater than the less than the first given threshold and/or second deviation
Two given thresholds then determine that the evaluating result of the same problems and/or the adjacent problem of the sequence is qualified.
Wherein, first given threshold and second given threshold are preset value.The same problems are to ask
In volume the problem of same type, such as: whether " gregarious " and " whether avoiding crowd ", above-mentioned two problems are relevant issues.One
The accurate questionnaire result of part, the answer of same problems should be relatively.Deviation between the answer of same problems
Can be calculated in the following way: the result value that such as questionnaire is each asked has " very between 0.0-8.0
Be not inconsistent ", " not being inconsistent substantially ", " uncertain ", " substantially conforming to " and " meeting very much " 5 quick options, corresponding option score
Respectively 0,2.0,4.0,6.0,8.0, for negative item, such as " whether avoiding crowd ", practical score are needed with " 8.0- choosing
Item score " conversion, if whether a questionnaire is above-mentioned " gregarious " and " whether avoiding crowd ", two topics all select " meeting very much ", then
Score 8.0,0.0 respectively, the first deviation between the two be | 8.0-0.0 |=8.0；And above-mentioned two topic of another questionnaire is all selected
" meeting ", then score 6.0,2.0 respectively.First deviation between the two is | 6.0-2.0 |=4.0.If the first setting threshold
Value is 7, then preceding a questionnaire is unqualified, and then a questionnaire is qualified, if the first given threshold is 3, two parts of questionnaires are not
It is qualified.
If the answer similarity of the problem of between sequence is adjacent is excessively high, it is likely that it is the random answer of answer person, therefore,
The second deviation between sequentially adjacent problem answers is greater than the second given threshold, then can be determined that related questionnaire is invalid.
Referring to Fig. 3, the questionnaire in the present embodiment is using continuous from the tradition selected topic using unlike simple discrete option
Option of the numberical range in conjunction with discrete index, value range maximin are consistent with traditional scale.In this way can
Enhance the accuracy of scale, and be able to maintain the readability of scale option, while being more advantageous to fine data analysis, intelligent Application
Correlation discovery in scene.
It is defeated by user property classification convolutional neural networks model for the ease of interpreting and utilizing user property classification results
The attributive classification of measurand out as a result, or (do not closed at this time by the attributive classification result of the measurand of questionnaire acquisition
Diagrid choosing), all text can be generated according to quantized data interpret report.Report provides solution according to dimension score stepping, and by stepping
The text information of reading.For example, it is open by [0.0-3.0), [3.0-5.0], (5.0-8.0] be divided into basic, normal, high third gear, interpret
Information respectively correspond evaluation sentence " your opening scoring is low, show you like with for no reason and simply mode is thought deeply.Other
It is pragmatical, pragmatic and conservative that people can be described as you ", " your open scoring is in average level, shows that you like passing
System, but be ready to attempt new things.Your idea was both remarkable or uncomplicated.For other people, you have seemingly received good religion
The people educated, but be not intellectual ", " your open scoring is high, shows your neophilia, multiplicity and variation.You are odd very well,
It is imaginative, have very much creativity ".It reports defeated by the dimension order of extropism, doing one's duty property, emotionality, opening and pleasant property
These evaluation sentences and measurand evaluating result chart out.
In one embodiment, the user property classification method of the invention based on convolutional neural networks, in use, such as
Fruit user property classification results deviation is excessive, then active user can be guided further to be tested and assessed by questionnaire, and according to two
The continuous iteration of attributive classification result of kind different modes updates user property classification convolutional neural networks model.Specifically: first use
The attributive classification of measurand is calculated by existing subscriber's attributive classification convolutional neural networks, and exports interpretation report.If with
Family (may be measurand either to the interested third party of measurand attributive classification) disagrees report content, anticipates
Taste attributive classification there may be relatively large deviations.It operates, can directly apply through questionnaire again if it is measurand
Assessment；If it is the third party, it can apply for generating questionnaire and test and assess again invitation, and invitation is sent to measurand, it
After measurand passes through invitation login and verifies identity by face recognition afterwards, questionnaire assessment can be carried out.Under two ways
The Psychological Evaluation result of the measurand obtained by questionnaire can be all saved.Thereafter, system accounting is calculated classifies from user property
The psychology of the attributive classification result of the measurand of convolutional neural networks model output and the measurand obtained by questionnaire is surveyed
The difference between result is commented, if being more than third given threshold, according to the Psychological Evaluation result for the measurand that the questionnaire obtains
Update convolutional neural networks model.
Third given threshold is preset value.
Referring to Fig. 4, the user property classification results that the questionnaire obtains according to Fig. 4 update convolutional neural networks model
Specific steps, comprising:
Step S21: by the psychological attribute related electronic data of the measurand and from user property classify convolutional Neural net
The attributive classification result of the measurand of network model output is put into alternative training set；
Step S22: the category of the measurand obtained from the output of user property classification convolutional neural networks model is obtained
Property classification results and questionnaire obtain measurand Psychological Evaluation result between difference；
Step S23: being more than the psychological attribute related electronic data of the measurand of third given threshold by the difference
Repetitive exercise collection is put into the Psychological Evaluation result of the measurand obtained by questionnaire；
Step S24: user property classification convolutional neural networks model is updated according to repetitive exercise collection.
The alternative training set is used to store the psychological attribute related electronic data of measurand and classifies from user property
The attributive classification of the measurand obtained in the output of convolutional neural networks model is as a result, to corresponding inspection in alternative training set
Result is surveyed to compare with the Psychological Evaluation result of the measurand obtained by questionnaire, if deviation is more than third given threshold,
Then think that deviation is excessive, the classification by user property classification volume machine network model is not accurate enough, is more than third by difference therefore
It the psychological attribute related electronic data of given threshold and is put into repeatedly with the Psychological Evaluation result of the measurand obtained by questionnaire
For training set, the repetitive exercise collection storage is for updating the measurand psychology of user property classification convolutional neural networks model
The Psychological Evaluation of attribute related electronic data and measurand is as a result, when repetitive exercise collection reaches certain amount, just by original training
Collection and the unification of repetitive exercise collection, and be trained again, the user property classification convolutional neural networks model updated.
In the present embodiment, in step S22, compare the measurand of user property classification convolutional neural networks model output
Attributive classification result and questionnaire obtain measurand Psychological Evaluation the result is that by comparing user extropism, and/or
Doing one's duty property and/or emotionality, and/or open and/or pleasant property.The difference calculates can be there are many calculation method, can
To be the mean difference, maximum difference or mean square deviation of each dimension.
User property classification method based on convolutional neural networks of the invention, it is easy to operate, it can largely save tested pair
As the time with appraiser, and convenient for being reused in the short time to same measurand, moreover it is possible to utilize NPU, GPU, TPU etc.
The higher-dimension computation capability of hardware.By using the psychological attribute related electronic data of user as input vector collection, by questionnaire
The Psychological Evaluation result of acquisition, to training user's attributive classification convolutional neural networks model, is used as the input of object vector collection
The attributive classification result that family attributive classification convolutional neural networks model obtains is more accurate, improves the efficiency of user property classification
And accuracy, more accurate user attribute data can be provided for the intelligent Application of internet, meanwhile, pass through the qualified heart of screening
Evaluating result is managed, the convolutional neural networks model trained is more accurate；Also using the biggish Psychological Evaluation result of deviation to
Family attributive classification convolutional neural networks model is iterated update, keeps user's attributive classification result more accurate.
The present invention also provides a kind of computer-readable storage medias, store computer program thereon, the computer program
The user property classification based on convolutional neural networks as described in any one of above-described embodiment is realized when being executed by processor
The step of method.
It wherein includes storage medium (the including but not limited to disk of program code that the present invention, which can be used in one or more,
Memory, CD-ROM, optical memory etc.) on the form of computer program product implemented.Computer-readable storage media packet
Permanent and non-permanent, removable and non-removable media is included, can be accomplished by any method or technique information storage.Letter
Breath can be computer readable instructions, data structure, the module of program or other data.The example packet of the storage medium of computer
Include but be not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM),
Other kinds of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory
(EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM), digital versatile disc
(DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetic storage devices or any other non-biography
Defeated medium, can be used for storage can be accessed by a computing device information.
The present invention also provides a kind of user property sorters, please refer to Fig. 5-7, and Fig. 5 is in an embodiment of the present invention
User property sorter overall structure diagram, user property sorter 50 include user property classification convolutional neural networks
Module 51, input module 52, output module 53 and feedback training module 54, the input module 52 receive the psychological attribute of user
Related electronic data, Psychological Evaluation result, detection type (Psychological Evaluation type)；The output module 53 exports the user of user
Attributive classification result；The user property classification convolutional neural networks module 51 is by a plurality of types of user properties classification convolution mind
It is composed in parallel through network, can be according to the psychological attribute related electronic data, while calculating the classification of multiple types user property
As a result；The training feedback module 54 by with the user property classify the one-to-one training feedback of convolutional neural networks model
Model is constituted, and the gap of attributive classification result and Psychological Evaluation result is calculated in training and to user property classification convolutional Neural
Network carries out feedback modifiers.As shown in fig. 6, input module 52 is inputted according to detection type (Psychological Evaluation type) in training
Corresponding convolutional neural networks and its loss function are opened, if the corresponding convolutional neural networks of detection type and its loss function
There is currently no then opened after dynamic increases newly by user property classification convolutional neural networks module 51 and feedback training module 54 respectively
Corresponding convolutional neural networks and its loss function are opened, then using the psychological attribute related electronic data of multiple users as input
Vector set is input in user property classification convolutional neural networks module 51, while the multiple use that will be obtained by questionnaire
The Psychological Evaluation result at family is input in feedback training module 54 as object vector collection, user property classification convolutional neural networks
After module 51 calculates user property classification results, the loss function being turned in training feedback module 54 passes through to output and mesh
Mark carries out costing bio disturbance and feeds back, training user's attributive classification convolutional neural networks module 51.
As shown in fig. 7, in use, input module 52 opens user property classification convolutional neural networks according to detection type
Designated model in module 51, and the psychological attribute related electronic data of measurand is passed into designated model, user property
In classification convolutional neural networks module 51 the part convolutional neural networks that are turned on calculate measurand attributive classification as a result,
It is integrated and is exported by output module 53.
The present invention also provides a kind of electronic equipment, referring to Fig. 8, Fig. 8 is electronic equipment knot in an embodiment of the present invention
Structure schematic diagram, electronic equipment 80 include controller 81 and memory 82, store computer program on the memory, the calculating
Such as the above-mentioned any one user as described in the examples based on convolutional neural networks is realized when machine program is executed by processor 81
The step of attributive classification method.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, (for example, directly will be without scale mould
The answer result of type processing itself, rather than treated Psychological Evaluation is as a result, as object vector training user's attributive classification
Convolutional neural networks, and dosage table model is processed into and the approximate attribute of Psychological Evaluation result again after neural network exports result
Classification results), these are all within the scope of protection of the present invention.
Claims (10)
1. a kind of user property classification method based on convolutional neural networks, which comprises the steps of:
Using the psychological attribute related electronic data of multiple users as input vector collection input user property classification convolutional Neural net
Network model；
Training feedback model is inputted using the Psychological Evaluation result of the multiple user obtained by questionnaire as object vector collection,
And update user property classification convolutional neural networks model；
The psychological attribute related electronic data of measurand input user property is classified convolutional neural networks model, and from user
The attributive classification result of attributive classification convolutional neural networks model output measurand.
2. the user property classification method according to claim 1 based on convolutional neural networks, it is characterised in that:
The psychology attribute related electronic data includes the facial picture of user and/or handwriting image and/or the user of user
The audio of speech and/or the video image of user's walking.
3. the user property classification method according to claim 2 based on convolutional neural networks, it is characterised in that:
Using the Psychological Evaluation result of the multiple user obtained by questionnaire as the input user property classification of object vector collection
Before convolutional neural networks model, further includes the steps that the qualified questionnaire evaluating result of screening, specifically includes:
The adjacent problem of sequence answers in the first deviation in acquisition questionnaire between the answer of same problems, and/or acquisition questionnaire
The second deviation between case；
If first deviation is greater than the second given threshold less than the first given threshold and/or second deviation,
Determine that the evaluating result of the same problems and/or the adjacent problem of the sequence is qualified.
4. the user property classification method according to claim 1 based on convolutional neural networks, which is characterized in that further include
Following steps:
If the attributive classification result of measurand that is obtained from the output of user property classification convolutional neural networks model and
Difference between the Psychological Evaluation result of the measurand obtained by questionnaire is more than third given threshold, then obtains the questionnaire
Measurand Psychological Evaluation result as object vector input, and update user property classification convolutional neural networks model.
5. the user property classification method according to claim 4 based on convolutional neural networks, it is characterised in that: ask this
The Psychological Evaluation result for rolling up the measurand obtained is inputted as object vector, and updates user property classification convolutional neural networks
Model specifically comprises the following steps:
It is exported by the psychological attribute related electronic data of the measurand and from user property classification convolutional neural networks model
The attributive classification result of measurand is put into alternative training set；
Obtain the attributive classification result of measurand that is obtained from the output of user property classification convolutional neural networks model with
Difference between the Psychological Evaluation result for the measurand that questionnaire obtains；
The difference is more than the psychological attribute related electronic data of the measurand of third given threshold and is obtained by questionnaire
The Psychological Evaluation result of the measurand taken is put into repetitive exercise collection；
User property classification convolutional neural networks model is updated according to repetitive exercise collection.
6. the user property classification method according to claim 1 based on convolutional neural networks, it is characterised in that:
The attributive classification result of the measurand includes all dimensions for trained Psychological Evaluation result, and can according to
It is changed in trained Psychological Evaluation type.
7. the user property classification method according to claim 3 based on convolutional neural networks, it is characterised in that:
The optional range of the answer of the questionnaire is continuous real-value range.
8. a kind of computer-readable storage media, stores computer program thereon, it is characterised in that: the computer program is located
It manages when device executes and realizes the user property classification side based on convolutional neural networks as claimed in any of claims 1 to 7 in one of claims
The step of method.
9. a kind of user property sorter characterized by comprising
Input module, for using the psychological attribute related electronic data of multiple users as input vector collection input user property point
Class convolutional neural networks model；
Feedback training module, the Psychological Evaluation result of the multiple user for that will be obtained by questionnaire is as object vector collection
Training feedback model is inputted, and updates user property classification convolutional neural networks model；
User property classification convolutional neural networks module, for the psychological attribute related electronic data of measurand to be inputted user
Attributive classification convolutional neural networks model, and divide from the attribute of user property classification convolutional neural networks model output measurand
Class result.
10. a kind of electronic equipment, including controller and memory, computer program is stored on the memory, feature exists
In: it is realized when the computer program is executed by processor and is based on convolutional Neural net as claimed in any of claims 1 to 7 in one of claims
The step of user property classification method of network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811246303.6A CN109086837A (en) | 2018-10-24 | 2018-10-24 | User property classification method, storage medium, device and electronic equipment based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811246303.6A CN109086837A (en) | 2018-10-24 | 2018-10-24 | User property classification method, storage medium, device and electronic equipment based on convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109086837A true CN109086837A (en) | 2018-12-25 |
Family
ID=64844071
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811246303.6A Pending CN109086837A (en) | 2018-10-24 | 2018-10-24 | User property classification method, storage medium, device and electronic equipment based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109086837A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110060772A (en) * | 2019-01-24 | 2019-07-26 | 暨南大学 | A kind of job psychograph character analysis method based on social networks |
| CN111460245A (en) * | 2019-01-22 | 2020-07-28 | 刘宏军 | Multi-dimensional crowd characteristic measuring method |
| WO2021003681A1 (en) * | 2019-07-09 | 2021-01-14 | LUI, Yat Wan | Method and system for neuropsychological performance test |
| CN112288686A (en) * | 2020-07-29 | 2021-01-29 | 深圳市智影医疗科技有限公司 | Model training method and device, electronic equipment and storage medium |
| CN113421175A (en) * | 2021-07-05 | 2021-09-21 | 赛飞特工程技术集团有限公司 | Capability test grading method and device |
| CN117574286A (en) * | 2024-01-11 | 2024-02-20 | 阿里健康科技(杭州)有限公司 | Method, device, equipment and storage medium for determining tag value |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120207375A1 (en) * | 2011-02-11 | 2012-08-16 | Fundacion Tecnalia Research & Innovation | Bioinspired System for Processing and Characterising Colour Attributes of a Digital Image |
| CN103106346A (en) * | 2013-02-25 | 2013-05-15 | 中山大学 | Character prediction system based on off-line writing picture division and identification |
| US20150186712A1 (en) * | 2013-02-08 | 2015-07-02 | Emotient | Collection of machine learning training data for expression recognition |
| US20160203729A1 (en) * | 2015-01-08 | 2016-07-14 | Happify, Inc. | Dynamic interaction system and method |
| CN106407935A (en) * | 2016-09-21 | 2017-02-15 | 俞大海 | Psychological test method based on face images and eye movement fixation information |
| CN106529402A (en) * | 2016-09-27 | 2017-03-22 | 中国科学院自动化研究所 | Multi-task learning convolutional neural network-based face attribute analysis method |
| CN106909896A (en) * | 2017-02-17 | 2017-06-30 | 竹间智能科技（上海）有限公司 | Man-machine interactive system and method for work based on character personality and interpersonal relationships identification |
| CN107799165A (en) * | 2017-09-18 | 2018-03-13 | 华南理工大学 | A kind of psychological assessment method based on virtual reality technology |
| CN107918780A (en) * | 2017-09-01 | 2018-04-17 | 中山大学 | A kind of clothes species and attributive classification method based on critical point detection |
| CN108171144A (en) * | 2017-12-26 | 2018-06-15 | 四川大学 | Information processing method, device, electronic equipment and storage medium |
-
2018
- 2018-10-24 CN CN201811246303.6A patent/CN109086837A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120207375A1 (en) * | 2011-02-11 | 2012-08-16 | Fundacion Tecnalia Research & Innovation | Bioinspired System for Processing and Characterising Colour Attributes of a Digital Image |
| US20150186712A1 (en) * | 2013-02-08 | 2015-07-02 | Emotient | Collection of machine learning training data for expression recognition |
| CN103106346A (en) * | 2013-02-25 | 2013-05-15 | 中山大学 | Character prediction system based on off-line writing picture division and identification |
| US20160203729A1 (en) * | 2015-01-08 | 2016-07-14 | Happify, Inc. | Dynamic interaction system and method |
| CN106407935A (en) * | 2016-09-21 | 2017-02-15 | 俞大海 | Psychological test method based on face images and eye movement fixation information |
| CN106529402A (en) * | 2016-09-27 | 2017-03-22 | 中国科学院自动化研究所 | Multi-task learning convolutional neural network-based face attribute analysis method |
| CN106909896A (en) * | 2017-02-17 | 2017-06-30 | 竹间智能科技（上海）有限公司 | Man-machine interactive system and method for work based on character personality and interpersonal relationships identification |
| CN107918780A (en) * | 2017-09-01 | 2018-04-17 | 中山大学 | A kind of clothes species and attributive classification method based on critical point detection |
| CN107799165A (en) * | 2017-09-18 | 2018-03-13 | 华南理工大学 | A kind of psychological assessment method based on virtual reality technology |
| CN108171144A (en) * | 2017-12-26 | 2018-06-15 | 四川大学 | Information processing method, device, electronic equipment and storage medium |
Non-Patent Citations (4)
| Title |
|---|
| F. GÜRPINAR 等: "Multimodal fusion of audio, scene, and face features for first impression estimation", 《2016 23RD INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)》 * |
| 中国内部审计协会 编: "《审计署内部审计科研课题研究报告2013-2014》", 28 February 2015, 中国时代经济出版社 * |
| 孙波 等: "智慧学习环境中基于面部表情的情感分析", 《现代远程教育研究》 * |
| 王伟军 编: "《大数据分析》", 30 April 2017, 重庆大学出版社 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111460245A (en) * | 2019-01-22 | 2020-07-28 | 刘宏军 | Multi-dimensional crowd characteristic measuring method |
| CN110060772A (en) * | 2019-01-24 | 2019-07-26 | 暨南大学 | A kind of job psychograph character analysis method based on social networks |
| CN110060772B (en) * | 2019-01-24 | 2022-07-01 | 暨南大学 | Occupational psychological character analysis method based on social network |
| WO2021003681A1 (en) * | 2019-07-09 | 2021-01-14 | LUI, Yat Wan | Method and system for neuropsychological performance test |
| CN112997166A (en) * | 2019-07-09 | 2021-06-18 | 晋斐德 | Method and system for neuropsychological performance testing |
| CN112288686A (en) * | 2020-07-29 | 2021-01-29 | 深圳市智影医疗科技有限公司 | Model training method and device, electronic equipment and storage medium |
| CN112288686B (en) * | 2020-07-29 | 2023-12-19 | 深圳市智影医疗科技有限公司 | Model training method and device, electronic equipment and storage medium |
| CN113421175A (en) * | 2021-07-05 | 2021-09-21 | 赛飞特工程技术集团有限公司 | Capability test grading method and device |
| CN117574286A (en) * | 2024-01-11 | 2024-02-20 | 阿里健康科技(杭州)有限公司 | Method, device, equipment and storage medium for determining tag value |
| CN117574286B (en) * | 2024-01-11 | 2024-05-24 | 阿里健康科技(杭州)有限公司 | Method, device, equipment and storage medium for determining tag value |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109086837A (en) | User property classification method, storage medium, device and electronic equipment based on convolutional neural networks | |
| Amrieh et al. | Mining educational data to predict student’s academic performance using ensemble methods | |
| Kenett et al. | Investigating the structure of semantic networks in low and high creative persons | |
| Anoopkumar et al. | A Review on Data Mining techniques and factors used in Educational Data Mining to predict student amelioration | |
| US11763174B2 (en) | Learning material recommendation method, learning material recommendation device, and learning material recommendation program | |
| Qi et al. | Evaluating on-line courses via reviews mining | |
| Wang et al. | Automatic generation of sentimental texts via mixture adversarial networks | |
| Tarun et al. | Generating licensure examination performance models using PART and JRip classifiers: A data mining application in education | |
| Shingari et al. | A review of applications of data mining techniques for prediction of students’ performance in higher education | |
| CN109636337A (en) | A kind of talent's base construction method and electronic equipment based on big data | |
| CN110070232B (en) | Method for multi-dimensional student score prediction by introducing teacher teaching style | |
| CN114154839A (en) | Course recommendation method based on online education platform data | |
| Wang et al. | A study of the evaluation metrics for generative images containing combinational creativity | |
| CN107578785A (en) | The continuous emotional feature analysis evaluation method of music based on Gamma distributional analysis | |
| Chaudhary et al. | Student future prediction using machine learning | |
| Alisherov et al. | Development of methods for assessing the performance of teachers using of TUIT-LMS data | |
| de Paula Santos et al. | “Speak well” or “complain” about your teacher: A contribution of education data mining in the evaluation of teaching practices | |
| Algur et al. | Educational data mining: RT and RF classification models for higher education professional courses | |
| Wang et al. | Construction and analysis of discrete system dynamic modeling of physical education teaching mode based on decision tree algorithm | |
| Oktavius et al. | Artificial Intelligence in entrepreneurial mindfulness using CRISP-DM Method | |
| Mustaffa et al. | Personality predictive analysis based on artificial neural network | |
| Minowa | Verification for generalizability and accuracy of a thinning-trees selection model with the ensemble learning algorithm and the cross-validation method | |
| Gustian et al. | Comparison of Classification Data Mining in Process Majors Students | |
| Wang et al. | Research on Student Achievement Analysis Method Based on Decision Tree Algorithm | |
| KR102624636B1 (en) | Self-feedback based questionnaire creation system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181225 |
|
| RJ01 | Rejection of invention patent application after publication |