CN108256022A - Talent evaluation model building method and personnel evaluation methods and system - Google Patents
Talent evaluation model building method and personnel evaluation methods and system Download PDFInfo
- Publication number
- CN108256022A CN108256022A CN201810023634.7A CN201810023634A CN108256022A CN 108256022 A CN108256022 A CN 108256022A CN 201810023634 A CN201810023634 A CN 201810023634A CN 108256022 A CN108256022 A CN 108256022A
- Authority
- CN
- China
- Prior art keywords
- talent
- evaluation
- sub
- model
- configuration
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
- G06F16/285—Clustering or classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Economics (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of talent evaluation model building method and the evaluation model based on structure carries out the system and method for talent evaluation.Talent's evaluation model construction method includes：Talent evaluation standard index system and mapping relations are configured；Talents information acquisition is carried out, and according to the talents information of acquisition and the mapping relations for the talent evaluation standard index system being configured, generate training set according to the talent evaluation standard index system of configuration；Evaluation model is configured, evaluation model is trained according to training set, generates model parameter.Pass through the method and system of the present invention, it can realize and unitize to talent evaluation standard, and evaluation model is trained based on unified standard, obtained evaluation model is built upon on the basis of unified standard, the ability level of the reflection talent that can be objective and accurate avoids based on different company standards evaluate and loses the bad of objective and fair.
Description
Technical field
The present invention relates to Talent Management technical field more particularly to a kind of talent evaluation model building method and based on the structure
The evaluation model that construction method trains carries out the method and system of talent evaluation.
Background technology
Current talent recommendation platform is normally applied the algorithm based on similarity and carries out position recommendation or post recommendation, such as
A kind of talent recommendation platform and Chinese invention patent application disclosed in Chinese invention patent application CN201610333304
A kind of talent ability model system and method based on big data, are all to get people disclosed in CN201610272595
After ability information, matched based on the similarity between employing unit and talents information, so as to fulfill talent recommendation.This side
Formula is merely able to realize that the degree of conformity between demand and talents information based on employing unit goes the ability level of the evaluation talent, and
The objective evaluation of the ability to the talent in itself is can not achieve, this is because：The ability of people is a kind of attribute of objective reality, without
Be with the demand between employing unit or same employing unit's difference position it is skimble-scamble, therefore, the need based on employing unit
The mode evaluated with the direct degree of conformity of talents information is sought, it cannot the objective ability level for embodying the talent.Therefore it provides
One, to realize the objective evaluation to the ability of the talent, can be had become by the index and evaluation method of entire corresponding field accreditation
For urgent problem to be solved in the industry.A kind of objective talent evaluation standard and method that can be generally applicable.
Invention content
One of the objects of the present invention is to provide a kind of personnel evaluation methods, are united by the talents arrangement for different field
One talent evaluation standard index system, and standard on data is carried out to the index system and is trained based on the index system
Go out model, realize the prediction to the grade mark of the talent, since the index system of configuration is can be suitable for corresponding industry to lead
The objective unified standard in domain, therefore, the talent's grade calculated by training pattern can objectively respond the ability level of the talent,
It avoids based on different company standards and carries out talent evaluation and lose the bad of objective and fair.
In order to realize the purpose, according to an aspect of the invention, there is provided a kind of talent evaluation model building method, is somebody's turn to do
Method includes：
Talent evaluation standard index system and mapping relations are configured；
Talents information acquisition is carried out according to the talent evaluation standard index system of configuration, and according to the talents information of acquisition and
The mapping relations of the talent evaluation standard index system of configuration generate training set；
Evaluation model is configured, evaluation model is trained according to training set, generates model parameter.It is in this manner it is possible to logical
The unified evaluation criteria system being suitble to for the corresponding talents arrangement of different field is crossed, and based on mapping relations, by the practical talent
Information is corresponding with the evaluation criteria system that the embodiment of the present invention is configured, and realizes and unitizes to talent evaluation standard, Zhi Houji
Evaluation model is trained in unified standard, obtains model parameter, obtained evaluation model is to comment foundation
The model of valency index system progress accurate evaluation, and the evaluation result that Utilization assessment model obtains, are just built upon unified standard
On the basis of, the ability level of the reflection talent that can be objective and accurate, avoiding based on different company standards evaluate has
Lose the bad of objective and fair.
In some embodiments, the evaluation model of configuration is softmax models.It can be realized based on Softmax models
Prediction to the probability for corresponding human just being assigned to some grade separation, due to acquisition talent's training examples concentrate input value and
Output valve is all known, thus when evaluation model is configured to softmax models, when being trained by given value, just
Output valve can be set to the maximization of known sample value, the model trained in this way is not only by adjusting model parameter
It realizes simply, and more tallies with the actual situation, accuracy rate higher.
In some embodiments, the talent evaluation standard index system of configuration includes essential attribute, technical merit, foreign language
Horizontal, commercial quality and talent's behavior top 5 factor, each element include corresponding sub- component attributes collection, wherein, substantially
The sub- component attributes of attribute include learning experiences and work experience；The sub- component attributes of technical merit include advanced procedures design,
Concept database and application, Principles of Operating System, computer network and technology, soft project, IT certifications and grade examination；Foreign language
Horizontal sub- component attributes include English, Japanese, French, Korean, Russian and Latin language；The sub- component attributes of commercial quality include
The known degree of the relevant laws such as communication capability, professional personality, intellectual property and team unity consciousness；The sub- element category of talent's behavior
Property include log in practice system frequency, online experiment frequency, online experiment report score, experiment quantity performed, experiment complete ratio
Rate, course quantity performed and course complete ratio；The mapping relations of the talent evaluation standard index system of configuration include actually adopting
The attribute vector of mapping relations and each talent's sample between the standardized value of the talents information of collection and each sub- component attributes with
Mapping relations between talent's grade separation standard.The index system covers all important indicators of IT talent, but also matches
The on-line study behavioral data of IT talent has been closed, has met actual requirement of the field to the talent, therefore, has been carried out based on the index
Grade separation and more objective, the reference value higher of evaluation.
In some embodiments, talents information acquisition, and root are carried out according to the talent evaluation standard index system of configuration
According to the mapping relations of the talent evaluation standard index system of the talents information and configuration of acquisition, generation training set includes：According to
Sub- component attributes in the talent evaluation standard index system put carry out information collection, obtain the information of each sub- component attributes；Root
According to the mapping relations between the talents information of the actual acquisition of configuration and the standardized value of each sub- component attributes, to each of acquisition
The information of sub- component attributes is standardized, and generates attribute vector；According to the attribute vector of each talent's sample of configuration and the talent
Mapping relations between grade separation standard determine the classification marker of each talent, generation training set { (x(1),y(1)) ..., (x(m),
y(m)), wherein, x(i)Represent the attribute vector of i-th of talent's sample, y(i)Represent the classification marker of i-th of talent's sample.Pass through
All sub- component attributes of configuration are all standardized, realize the unification of minimum evaluation element, and to talent's grade
Classification marker is also to establish on the basis of standardized value, and therefore, the standard of grade separation is also unified.Further, since instruction
It is based on unitized talents information to practice sample, thus the training examples collection established is also unified, in unified training examples
The model training carried out on collection so that evaluation model can better adapt to and coordinate the index system that the embodiment of the present invention is configured
System that is, so that the evaluation model trained is the unified evaluation and foreca carried out under the evaluation criteria system of the present invention, is based on
Evaluation criteria system unitizes and objectifies, and obtained evaluation and foreca result is also more objective, more tallies with the actual situation.
According to another aspect of the present invention, a kind of personnel evaluation methods are additionally provided, this method is first by aforementioned
Talent evaluation model building method trains evaluation model, later according to the grade of the talent evaluation model prediction talent constructed
Classification.The accurate evaluation for unitizing, objectifying to the talent is realized according to the evaluation model that preceding method trains, therefore,
Based on talent's grade separation prediction that the model carries out, unified evaluation and classification to the different field talent are realized, it is more accurate
Really, reference value higher.
In some embodiments, talent evaluation model is softmax models, and model formation is：
Wherein,For model parameter, T is the transposition symbol in linear algebra,x(i)Represent i-th of talent's sample
This attribute vector,Perseverance is 1,J=1,2 ..., n represents j-th of sub- component attributes variable of i-th of talent's sample
Value, the number of the optional value set of y are labeled as k.By the model, the sample that can be turned out talents with the prediction of simple, intuitive is evaluated
For the probability of each grade, the probability of each grade is be evaluated as by the talent, can more accurately embody the ability of the talent
It is horizontal.And predicted by way of probability, it can also have gained some understanding to the ability development trend of the talent, it can be from whole
Reflect the integration capability and quality of the corresponding talent on body, more meet talent's characteristic attribute, compare the evaluation for only carrying out single result
Mode, the algorithm are more suitable for the application scenarios of this complexity of talent ability.
In some embodiments, included according to the grade separation of the trained softmax model prediction talents：It obtains to be evaluated
The sub- element vector set x of the talent of valency(i), according to the predicted value y of model generation talent grade separation label trained(i)∧'s
The probability of value；The y acquired(i)∧All values in maximum probability label of the value as the talent to be evaluated.
After training evaluation mould, you can the prediction of grade separation is carried out to unknown talent's sample, is carried out by softmax models pre-
It surveys, can cross and reflect distribution probability of the talent's sample in each classification grade of classification grade standard, and the classification of maximum probability
Grade, the current ability for just more meeting the talent is horizontal, therefore, is precisely predicted when needing the current ability level to the talent
When, in all values acquired, the value of maximum probability, and when needing comprehensive condition and ability development to the talent
Trend judge, can refer to the probability distribution situation of all values, very flexibly, more can accurately embody the reality of the talent
Capabilities might.
In some embodiments, the y acquired(i)∧All values in the value of maximum probability be by following public affairs
Formula is realized：
Wherein,It represents to enableObtain the j of maximum value.It can be quickly and easily automatic by the formula
The value of maximum probability is got, that is, directly obtains the classification marker for more meeting the current ability level of the talent, is not required to very important person
To judge result, efficiency is improved, suitable for situation about being evaluated simultaneously a large amount of talent.
According to a further aspect of the invention, a kind of talent evaluation system is additionally provided, which matches including evaluation criterion
Module, evaluation criterion conversion module, model training module and grade forecast module are put, wherein, evaluation criterion configuration module is used for
Talent evaluation standard index system and the storage of its mapping relations and configuration evaluation model storage is configured；Evaluation criterion converts mould
Block is used to map talent's sample information according to the mapping relations of the talent evaluation standard index system of configuration, generates the talent
Attribute vector and the label storage of talent's grade separation；Model training module is used for according to talent's attribute vector and talent's grade separation
Label generation training set, and the evaluation model of configuration is trained according to training set, model parameter is generated, exports training mould
Type；The grade separation predicted value that grade forecast module is used to generate the talent to be evaluated according to training pattern exports.Pass through the present invention
The talent evaluation system that embodiment provides, you can it realizes the automatic Evaluation to the talent, does not need to manually be calculated, it is efficient.
Furthermore, it is possible to which talent evaluation standard index system and mapping relations are configured according to demand, to adapt to the requirement of different field, it is applicable in
Range is wide.And merit rating and prediction are carried out to the talent in same field based on unified standard, as a result more want reference value.
In some embodiments, wherein, which further includes information acquisition module, for being referred to according to talent evaluation standard
Mark system acquires corresponding talents information, obtains the storage of talent's sample information.Pass through automatic collection talents information, it is possible to reduce people
Work operates, such as the operation of typing talents information, improves efficiency, reduces typing error rate, more convenient.
Description of the drawings
Fig. 1 is the flow diagram of the personnel evaluation methods of one embodiment of the present invention；
Fig. 2 is the circuit theory schematic diagram of the talent evaluation system of one embodiment of the present invention.
Specific embodiment
Embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.
The present invention is intended to provide a kind of generally applicable personnel evaluation methods, core idea is by for corresponding field
The unified evaluation criterion index system of talents arrangement, and quantize to the evaluation criterion index system of configuration, later, base
Training set is obtained in evaluation criterion index system and numeralization standard acquisition, and passes through training set and selected model is instructed
Practice, obtain evaluation model.Then, it is possible to which the evaluation model that application obtains carries out analysis calculating to unknown example, obtains the talent
Classification grade.Due to being directed to the talent in same field, the evaluation criterion index system and numeralization standard of configuration be it is unified,
Therefore it is objective to the evaluation of the talent in same field, avoids based on different company standards evaluate and lose objective public affairs
Positive is bad.Fig. 1 shows a kind of method evaluated based on softmax models the talent, in embodiments of the present invention, choosing
Evaluation model is softmax models, and evaluation object is evaluates IT talent, as shown in Figure 1, in the embodiment of the present invention
Middle method can be implemented as：
Step S101：Determine IT talent's evaluation criterion index system and its correlator element.
In the embodiment of the present invention, according to IT industries to the composite request and actual conditions of the talent, determining IT talent evaluation
Standard index system is mainly made of top 5 factor, can be respectively " essential attribute ", " technical merit ", " L proficiency ", " quotient
Business quality " and " talent's behavior ".Specify corresponding sub- element collection as follows for each element：
Sub- element for " essential attribute " distribution is to include " learning experiences " and " work experience "；It is distributed for " technical merit "
Sub- element be include " advanced procedures design ", " concept database and application ", " Principles of Operating System ", " computer network and
Technology ", " soft project ", " IT certifications and grade examination "；Sub- element for " L proficiency " distribution is to include " English ", " day
Language ", " French ", " Korean ", " Russian ", " Latin language " etc.；For " commercial quality " distribution sub- element be include " communication capability ",
" professional personality ", " the known degree of the relevant laws such as intellectual property ", " team unity consciousness "；Son for " talent's behavior " distribution will
Element be include " logging in practice system frequency ", " online experiment frequency ", " online experiment report score ", " experiment quantity performed ",
" ratio is completed in experiment ", " course quantity performed ", " course completion ratio ".Wherein, " talent's behavior " is primarily referred to as job hunter's
Behavioral data, the embodiment of the present invention are illustrated so that behavior data is the behavioral datas of the on-line study of job hunter as an example,
Behavior data can also be other sub- element collection in other embodiment, as long as the behavioral data of job hunter can be characterized i.e.
Can, such as it is mating with the respective behavior of corresponding on-line study system.
Step S102：Determine the attribute number value standard of each sub- element of IT talent's evaluation criterion index system.
To the sub- element collection that the sub- element of each element determined in step S101 forms, examined by investigation and reference grade
The modes such as the examination index of examination, establish mapping function, to realize the numeralization to the sub- component attributes of talent's example.Wherein, it investigates
It is primarily referred to as carrying out data collection, obtains the different corresponding various grade examinations of sub- component attributes and its index of taking an examination, such as
Data information is obtained in a manner that web crawlers carries out data collection, by inquiry questionnaire, is obtained by specific APP clients
Take user data information etc..And mapping function is established to realize the numeralization to the sub- component attributes of talent's example, it is primarily referred to as building
The score of the quasi- grade examination of day-mark is to the mapping of the numerical value of sub- component attributes, the type and its corresponding examination that standard class is taken an examination
Score is obtained by data collection, and the numerical value of sub- component attributes can be set as a unified standard on data according to demand,
Such as someone achievement 60 of TOEFL examinations is divided hereinafter, can then set corresponding English proficiency as 1 point of (i.e. sub- component attributes
Numerical value be mapped as 1), the achievement of TOEFL examinations is 61-70 point, then can set corresponding English proficiency as 2 points ...,
TOEFL total marks of the examination are 105 points or more, then can set corresponding English proficiency as 5 points, and the achievement of IELTS is 7 points or more
Then corresponding English proficiency is set as 5 and grades.In this way, by investigating the modes such as the examination index taken an examination with reference grade, it is
Different sub- component attributes configuration standard numerical value, it is possible to which realization avoids different Computer Examination Evaluating Systems to sub- component attributes value
It influences, realizes the unification of the evaluation criterion to the sub- element of configuration, and the setting of the specific mapping value of each sub- component attributes,
It can arbitrarily set according to demand, as long as so that corresponding sub- component attributes are based on unified standard, the present invention is implemented
Example does not limit specific mapping value.
It should be noted that here establish standard class examination score and sub- component attributes numerical value between reflect
It penetrates, can be the information got according to investigation and the numeralization standard set, establish matching rule and stored, to obtain
To certain job hunter sub- component attributes value when, by matching rule store mapping relations, by the sub- element information got turn
It turns to the sub- element numerical value of standard or according to mapping relations between the two, such as is examined based on the grade that investigation obtains
Mapping relations between the examination target time value of examination and determining sub- element standard figures, select suitable mathe-matical map function
Realize mapping between the two, the specific implementation that the embodiment of the present invention realizes the numeralization of sub- component attributes to establishing mapping function
Mode is not limited, and the sub- element information of the talent of acquisition is mapped as unified standard figures as long as can realize.
Step S103：According to IT talent's evaluation criterion, talents information is acquired, stamps tag along sort.
After the numeralization standard of above-mentioned evaluation criterion index system and sub- component attributes is configured with for IT talent, it is possible to
Sub- element collection in These parameters system, is acquired the information of IT talent, such as pass through web crawlers, questionnaire
Information collection is carried out, and according to collected information and each sub- component attributes of configuration with modes such as dedicated app clients
Numeralization standard, generate training set.Wherein, according to collected information and the numeralization mark of each sub- component attributes of configuration
Quasi- generation training set process be：By all sub- component attributes variables in collected IT talent's evaluation criterion index system
The mapping that value passes through in step S102 is converted into unified numerical value, and the numerical value set of the sub- component attributes after conversion is labeled as
X, missing values are replaced by the median of the variable, wherein, missing values refer in data acquisition since data source is inconsistent
Or the reasons such as data are imperfect and caused by do not get the situation of the corresponding numerical value of sub- component attributes of the sample, and middle position
It is several, refer to the intermediate value of the set of all effective values of the collected sub- component attributes, for example, acquiring four talent's samples
This, the age of four people is respectively 18, null, 22,23, and missing values are exactly null, and median is exactly 22.By the son in set
The number of component attributes variable is labeled as n, and the corresponding talent's grade separation of each sub- element is labeled as y, the criteria for classification root of y
Determine that (standard can refer to existing classification in human resource system according to the practical job hunting situation of IT talent, wages treatment etc.
Standard), such as y can be configured to grade 1,2,3,4,5,1 and represent best, 5 represent the worst talent, are as a result,
Can training set be generated according to the sub- element variable of acquisition and the corresponding talent's grade separation of sub- element variable, you can by talents information
Training set be labeled as { (x(1),y(1)) ..., (x(m),y(m)), whereinRepresent x(i)It is n+1 dimensional vectors, x(i)Table
Show the attribute vector (i.e. the set of the corresponding collected sub- element information of i-th of talent's sample) of i-th of talent's sample,
Perseverance is 1,(j=1,2 ..., n) represents the value of j-th of sub- component attributes variable in i-th of talent's sample, and j represents jth
A sub- component attributes；Y (such as is being divided into the example that includes grade 1,2,3,4,5 by the number of the optional value set of y labeled as k
In son, the selectable value of y is just 1,2,3,4,5, and the number k of the set formed is just for 5).Wherein, by x(i)Be defined as n+1 tie up to
Amount, andPerseverance is 1, in this manner it is possible to increase by one on the basis of the linear combination of all properties of x not by the attribute shadow of x
Loud bias improves the accuracy rate that talent's grade separation prediction is carried out by model.
Step S104：Training softmax models.
Training refers to the process of that according to given data searching model parameter the mapping finally searched, which is referred to as training, to be come
Model.And the essence of Softmax models is exactly to tie up the arbitrary real vector V1 compressions (mapping) of a n dimension into another k
Real vector V2, wherein, each element value in another real vector of generation is between (0,1).The present invention
Embodiment is based on the training sample set determined in step S103, softmax models is trained, for the grade to the talent
Classification is predicted.Wherein, shown in the softmax models such as formula (1) of setting, model parameter isThe formula represents the talent in x
In the case of attribute, it is according to parameter valueModel, the probability for this talent being assigned to corresponding grade y can be predicted.
Due to the attribute vector x of the training sample set in known steps S103(i)Y corresponding with each attribute vector each trains sample
The talent of this concentration has corresponded to y labels, for example it is 2 that training sample, which concentrates the corresponding y of some sample, i.e., in training sample set
In y to determine value known to one, therefore based on the attribute vector x in sample set(i)With y can to the model of formula (1) into
Row training, to adjust model parameterObtain probability so that y obtains mutually deserved determining value such as 2 and be the bigger the better, and y obtain other can
The probability of choosing value is the smaller the better, thus train model parameter can so that practical unknown sample to obtain prediction result more smart
It is accurate.
It can be seen that being to adjust model parameter using the purpose of training set training pattern, step up the accuracy rate of model.
In the concrete realization, training technique of the prior art can be selected to be trained model, such as can be declined using gradient
Model parameter in the technique drills such as method or L-BFGS algorithms formula (1).In a preferred embodiment, also by minimizing cost letter
It counts to calculate gradient, carries out the optimization of model parameter, be shown below, obtained minimum cost function J (θ) is：
Wherein, 1 { expression formula } is exponential function, and IF expression is true, then 1 { expression formula } is equal to 1, otherwise 1 { expression
Formula } equal to 0.Model parameter can be optimized by the minimum cost function.Wherein, the specific optimization of optimization algorithm
Process and principle are referred to the prior art, the embodiment of the present invention to this without repeating, it should be understood by those skilled in the art that
, the present invention can select a variety of existing optimization algorithms to optimize model parameter, and minimize cost function progress
Optimization is a specific example, is not intended as the limitation to the optimization algorithm of selection.
Step S105：Based on softmax models, the grade separation of IT talent is predicted.
After softmax models are trained, i.e., after the model of formula (1), each height for obtaining IT talent to be evaluated will
The variable-value of sub- element is substituted into the softmax models of formula (1), i.e., by the corresponding variate-value (i.e. unknown talent's sample) of element
The probability of all values of the corresponding classification of qualified personnel standard y of sub- element collection of the talent can be obtained, is taken as shown in formula (1)
The y that softmax models obtain(i)∧All values in maximum probability label (i.e. grade separation of the value as the IT talent
Label), be shown below for：
Wherein,It represents to enableThe j of maximum value is obtained, and obtained j has corresponded to criteria for classification y's
The number of selectable value can more be met the criteria for classification value of the IT talent by j as a result,.Since softmax models obtain
To be maximum probability in all values value, therefore, label of the value as IT talent more meets objective reality, right
The grade classification of the talent is more precisely reliable, has higher reference value.And thus it can also be seen that passing through softmax moulds
Type is predicted, is realized very simple.
Fig. 2 schematically shows the frame structure of the talent evaluation system under a kind of realization method of the present invention.Such as Fig. 2 institutes
Show, which includes memory module 20, evaluation criterion configuration module 21, evaluation criterion conversion module 22, model training module 23
With grade forecast module 25.Wherein, evaluation criterion configuration module 21 is used to that talent evaluation standard index system to be configured and it is reflected
The relationship of penetrating is stored to memory module 20, and is also configured evaluation model and stored to memory module 20, wherein, talent evaluation standard refers to
Mark system includes fundamental, is corresponded in each fundamental and includes multiple sub- component attributes, mapping relations include will be each
The corresponding social information of sub- component attributes (the corresponding various evaluation informations of each sub- component attributes i.e. in reality) is mapped as standard
The numerical value of change and the sub- component attributes DUAL PROBLEMS OF VECTOR MAPPING taken after standardized value of each talent is met into established standards for one
The talent's grade separation label.Evaluation criterion conversion module 22 is used to generate the sub- element of each talent's sample according to mapping relations
Standardized value vector set and the classification of qualified personnel label of attribute are stored to memory module 20, wherein, the sub- element of each talent
Attribute vector collection has corresponded to matching talent's grade separation label, that is, forms one to value.Model training module
23 are used to generate instruction to value according to what the attribute vector collection and talent's grade separation formed in evaluation criterion conversion module 22 marked
Practice collection, and generation model parameter is trained evaluation model according to training set, export trained evaluation model.Grade forecast
The grade separation predicted value that module 25 is used to generate the talent to be evaluated according to training pattern (i.e. trained evaluation model) exports.
Wherein, as shown in Fig. 2, the system further includes information acquisition module 24, for according to talent evaluation standard index system acquisition phase
The talents information answered, so as to obtain the storage of talent's sample information.In the concrete realization, user or administrator can be marked by evaluating
The user interface setting talent evaluation standard index system and its mapping relations and evaluation model of quasi- configuration module 21, and by evaluating
Standard configuration module 21 stores the configuration information received, such as passes through database purchase.Later, evaluation criterion conversion module
22 talent's sample information for acquiring acquisition information acquisition module 24, conversion is carried out to talent's sample information according to mapping relations will
It is mapped as standard figures storage, that is, the gathered data corresponding with sub- component attributes of each talent's sample is obtained, according to setting
Mapping relations, be mapped as sub- component attributes standard figures, so as to obtain the sub- element vector set of each talent's sample, later
Further according to mapping relations and the standard figures of sub- component attributes, obtain the corresponding talent's grade separation label of talent's sample and deposit
Storage.Later, model training module 23 will be according to the vector set of talent's sample and the corresponding grade evaluation mark of each talent's sample
Quasi- generation training set, is trained with being based on model (such as softmax models), model parameter is trained, so as to be trained
Model.Finally, when evaluating some talent, the sub- element information that need to only obtain the talent to be evaluated (such as passes through
Data acquisition module is inputted by user), it is possible to the module for training the sub- element information substitution of the talent to be evaluated,
Realize the prediction to the classification grade of the talent.When being trained based on softmax models, the prediction result of output is the people
The probability of all values of the corresponding grade separation of sub- element information of, at this time by obtaining most probable value, you can obtain
The grade mark of the talent.By taking model is softmax models as an example, the specific implementation of each module in system of the embodiment of the present invention
Process can refer to the narration of method part above, and details are not described herein.Wherein, in a particular embodiment, the talent can be IT people
Ability or the talent in other industry field, by taking IT talent as an example, talent evaluation standard index system can be by belonging to substantially
Property, technical merit, L proficiency, commercial quality and talent's behavior top 5 factor composition, each element include corresponding son
Element collection, the sub- element for " essential attribute " distribution is to include " learning experiences " and " work experience "；It is distributed for " technical merit "
Sub- element be include " advanced procedures design ", " concept database and application ", " Principles of Operating System ", " computer network and
Technology ", " soft project ", " IT certifications and grade examination "；Sub- element for " L proficiency " distribution is to include " English ", " day
Language ", " French ", " Korean ", " Russian ", " Latin language " etc.；For " commercial quality " distribution sub- element be include " communication capability ",
" professional personality ", " the known degree of the relevant laws such as intellectual property ", " team unity consciousness "；Son for " talent's behavior " distribution will
Element be include " logging in practice system frequency ", " online experiment frequency ", " online experiment report score ", " experiment quantity performed ",
" ratio is completed in experiment ", " course quantity performed ", " course completion ratio ".
Above-described is only some embodiments of the present invention.For those of ordinary skill in the art, not
Under the premise of being detached from the invention design, various modifications and improvements can be made, these belong to the protection model of the present invention
It encloses.
Claims (10)
1. talent evaluation model building method, which is characterized in that including：
Talent evaluation standard index system and mapping relations are configured；
Talents information acquisition, and the talents information according to acquisition and configuration are carried out according to the talent evaluation standard index system of configuration
Talent evaluation standard index system mapping relations, generate training set；
Evaluation model is configured, evaluation model is trained according to training set, generates model parameter.
2. according to the method described in claim 1, wherein, the evaluation model of configuration is softmax models.
3. according to the method described in claim 2, wherein, the talent evaluation standard index system of configuration includes basic belong to
Property, technical merit, L proficiency, commercial quality and talent's behavior top 5 factor, each element include corresponding sub- element
Property set, wherein,
The sub- component attributes of the essential attribute include learning experiences and work experience；
The sub- component attributes of the technical merit include advanced procedures design, concept database and application, Principles of Operating System, meter
Calculation machine network and technology, soft project, IT certifications and grade examination；
The sub- component attributes of the L proficiency include English, Japanese, French, Korean, Russian and Latin language；
The sub- component attributes of the commercial affairs quality include the known degree of the relevant laws such as communication capability, professional personality, intellectual property
Realize with team unity；
The sub- component attributes of talent's behavior are reported including logging in practice system frequency, online experiment frequency, online experiment
Divide, experiment quantity performed, experiment completion ratio, course quantity performed and course complete ratio；
The mapping relations of the talent evaluation standard index system of configuration include the talents information of actual acquisition and each sub- component attributes
Standardized value between mapping relations and the attribute vector of each talent's sample and talent's grade separation standard between mapping
Relationship.
It is 4. described that the talent is carried out according to the talent evaluation standard index system of configuration according to the method described in claim 3, wherein
Information collection, and according to the talents information of acquisition and the mapping relations for the talent evaluation standard index system being configured, generation training
Collection includes：
Sub- component attributes in the talent evaluation standard index system of configuration carry out information collection, obtain each sub- component attributes
Information；
According to the mapping relations between the standardized value of the talents information of the actual acquisition of configuration and each sub- component attributes, to obtaining
The information of each sub- component attributes taken is standardized, and generates attribute vector；
According to the mapping relations between the attribute vector of each talent's sample of configuration and talent's grade separation standard, each talent is determined
Classification marker, generation training set { (x(1),y(1)) ..., (x(m),y(m)), wherein, x(i)Represent the attribute of i-th of talent's sample
Vector, y(i)Represent the classification marker of i-th of talent's sample.
5. personnel evaluation methods, which is characterized in that including：
Build talent evaluation model；
According to the grade separation of the talent evaluation model prediction talent constructed；
Wherein, the method for structure talent evaluation model is by Claims 1-4 any one of them talent evaluation model construction
Method is realized.
6. according to the method described in claim 5, wherein, talent evaluation model is softmax models, model formation is：
Wherein, θ is model parameter, and T is transposition symbol,x(i)Represent the attribute vector of i-th of talent's sample,
Perseverance is 1,J=1,2 ..., n represent the value of j-th of sub- component attributes variable of i-th of talent's sample, the selectable value collection of y
The number of conjunction is labeled as k.
7. according to the method described in claim 6, wherein, the ranking score according to the trained softmax model prediction talents
Class includes：
Obtain the sub- element vector set x of the talent to be evaluated(i), marked according to the model generation talent grade separation trained
Predicted value y(i)^Value probability；
The y acquired(i)^All values in maximum probability label of the value as the talent to be evaluated.
8. according to the method described in claim 7, wherein, the y acquired(i)^All values in the value of maximum probability be
It is realized by following formula：
Wherein,It represents to enableObtain the j of maximum value.
9. talent evaluation system, which is characterized in that including evaluation criterion configuration module, evaluation criterion conversion module, model training
Module and grade forecast module, wherein,
The evaluation criterion configuration module is stored and is configured for talent evaluation standard index system and its mapping relations to be configured
Evaluation model stores；
The evaluation criterion conversion module is used for the mapping relations according to the talent evaluation standard index system of configuration to talent's sample
Example information is mapped, and generates talent's attribute vector and the label storage of talent's grade separation；
The model training module is used for according to talent's attribute vector and talent's grade separation label generation training set, and according to instruction
The evaluation model for practicing set pair configuration is trained, and generates model parameter, exports training pattern；
The grade separation predicted value that the grade forecast module is used to generate the talent to be evaluated according to training pattern exports.
10. system according to claim 9, wherein, information acquisition module is further included, for referring to according to talent evaluation standard
Mark system acquires corresponding talents information, obtains the storage of talent's sample information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810023634.7A CN108256022A (en) | 2018-01-10 | 2018-01-10 | Talent evaluation model building method and personnel evaluation methods and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810023634.7A CN108256022A (en) | 2018-01-10 | 2018-01-10 | Talent evaluation model building method and personnel evaluation methods and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108256022A true CN108256022A (en) | 2018-07-06 |
Family
ID=62724820
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810023634.7A Pending CN108256022A (en) | 2018-01-10 | 2018-01-10 | Talent evaluation model building method and personnel evaluation methods and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108256022A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109523056A (en) * | 2018-10-12 | 2019-03-26 | 中国平安人寿保险股份有限公司 | Object ability classification prediction technique and device, electronic equipment, storage medium |
| CN110851582A (en) * | 2018-07-25 | 2020-02-28 | 北京京东尚科信息技术有限公司 | Text processing method and system, computer system and computer readable storage medium |
| CN110909970A (en) * | 2018-09-17 | 2020-03-24 | 北京京东金融科技控股有限公司 | Credit scoring method and device |
| CN111369131A (en) * | 2020-02-28 | 2020-07-03 | 深圳前海微众银行股份有限公司 | Talent assessment method, device and equipment based on AI and readable storage medium |
| CN111967729A (en) * | 2020-07-28 | 2020-11-20 | 兰笺(苏州)科技有限公司 | Industrialized personnel portrait evaluation method based on data mining |
| CN113342793A (en) * | 2021-06-18 | 2021-09-03 | 立信(重庆)数据科技股份有限公司 | Investigation data standardization method and system |
| CN116011979A (en) * | 2022-12-07 | 2023-04-25 | 深圳今日人才信息科技有限公司 | Method for constructing talent competitiveness index based on complex network |
| CN117787799A (en) * | 2023-12-28 | 2024-03-29 | 国家电投集团科学技术研究院有限公司 | Talent evaluation method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150199646A1 (en) * | 2014-01-16 | 2015-07-16 | Hirevue, Inc. | Model-assisted evaluation and intelligent interview feedback |
| CN105160498A (en) * | 2015-10-21 | 2015-12-16 | 北京普猎创新网络科技有限公司 | Personal value calculation method based on big data |
| CN105787639A (en) * | 2016-02-03 | 2016-07-20 | 北京云太科技有限公司 | Artificial-intelligence-based talent big data quantization precise matching method and apparatus |
| CN105931116A (en) * | 2016-04-20 | 2016-09-07 | 帮帮智信（北京）教育投资有限公司 | Automated credit scoring system and method based on depth learning mechanism |
-
2018
- 2018-01-10 CN CN201810023634.7A patent/CN108256022A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150199646A1 (en) * | 2014-01-16 | 2015-07-16 | Hirevue, Inc. | Model-assisted evaluation and intelligent interview feedback |
| CN105160498A (en) * | 2015-10-21 | 2015-12-16 | 北京普猎创新网络科技有限公司 | Personal value calculation method based on big data |
| CN105787639A (en) * | 2016-02-03 | 2016-07-20 | 北京云太科技有限公司 | Artificial-intelligence-based talent big data quantization precise matching method and apparatus |
| CN105931116A (en) * | 2016-04-20 | 2016-09-07 | 帮帮智信（北京）教育投资有限公司 | Automated credit scoring system and method based on depth learning mechanism |
Non-Patent Citations (1)
| Title |
|---|
| ALEXANDERKUN: "softmax函数详解", 《HTTPS://WWW.CNBLOGS.COM/ALEXANDERKUN/P/8098781.HTML》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110851582A (en) * | 2018-07-25 | 2020-02-28 | 北京京东尚科信息技术有限公司 | Text processing method and system, computer system and computer readable storage medium |
| CN110909970A (en) * | 2018-09-17 | 2020-03-24 | 北京京东金融科技控股有限公司 | Credit scoring method and device |
| CN109523056A (en) * | 2018-10-12 | 2019-03-26 | 中国平安人寿保险股份有限公司 | Object ability classification prediction technique and device, electronic equipment, storage medium |
| CN109523056B (en) * | 2018-10-12 | 2023-11-07 | 中国平安人寿保险股份有限公司 | Object capability classification prediction method and device, electronic equipment and storage medium |
| CN111369131A (en) * | 2020-02-28 | 2020-07-03 | 深圳前海微众银行股份有限公司 | Talent assessment method, device and equipment based on AI and readable storage medium |
| CN111369131B (en) * | 2020-02-28 | 2024-04-30 | 深圳前海微众银行股份有限公司 | Talent assessment method, device and equipment based on AI (advanced technology attachment), and readable storage medium |
| CN111967729A (en) * | 2020-07-28 | 2020-11-20 | 兰笺(苏州)科技有限公司 | Industrialized personnel portrait evaluation method based on data mining |
| CN113342793A (en) * | 2021-06-18 | 2021-09-03 | 立信(重庆)数据科技股份有限公司 | Investigation data standardization method and system |
| CN113342793B (en) * | 2021-06-18 | 2023-04-07 | 立信(重庆)数据科技股份有限公司 | Research data standardization method and system |
| CN116011979A (en) * | 2022-12-07 | 2023-04-25 | 深圳今日人才信息科技有限公司 | Method for constructing talent competitiveness index based on complex network |
| CN117787799A (en) * | 2023-12-28 | 2024-03-29 | 国家电投集团科学技术研究院有限公司 | Talent evaluation method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108256022A (en) | Talent evaluation model building method and personnel evaluation methods and system | |
| Lock Lee et al. | Visualising and measuring intellectual capital in capital markets: a research method | |
| Braun et al. | Predicting stock market behavior through rule induction: an application of the learning‐from‐example approach | |
| CN116823578A (en) | Intelligent city planning system and method based on big data analysis | |
| Asanbe et al. | Teachers’ performance evaluation in higher educational institution using data mining technique | |
| Fang | Intelligent online English teaching system based on SVM algorithm and complex network | |
| CN110378206B (en) | Intelligent image examination system and method | |
| Brito et al. | A conceptual comparison between discrete and continuous simulation to motivate the hybrid simulation methodology | |
| CN109934469A (en) | Based on the heterologous power failure susceptibility method for early warning and device for intersecting regression analysis | |
| CN109933783A (en) | A kind of essence of a contract method of non-performing asset operation field | |
| CN102591929A (en) | Library data processing system and data processing method thereof | |
| Dragomir et al. | Does corporate governance improve integrated reporting quality? A meta-analytical investigation | |
| Azeroual et al. | Data quality strategy selection in CRIS: using a hybrid method of SWOT and BWM | |
| CN109255389A (en) | A kind of equipment evaluation method, device, equipment and readable storage medium storing program for executing | |
| CN106503050B (en) | Method and system for recommending reading articles based on big data | |
| Ye | The decision tree classification and its application research in personnel management | |
| Zhou et al. | Representation and analysis of time-series data via deep embedding and visual exploration | |
| CN116431828A (en) | Construction method of power grid center data asset knowledge graph database constructed based on neural network technology | |
| CN115984053A (en) | Test question recommendation method and device, electronic equipment and storage medium | |
| Li | Improvement design of functional modules of college english network teaching system based on decision tree algorithm | |
| Pei | Construction and application of talent evaluation model based on nonlinear hierarchical optimization neural network | |
| Krishnan et al. | Performance measurement link between the balanced scorecard dimensions: an empirical study of the manufacturing sector in Malaysia | |
| Srimani et al. | Regression model for Edu-data in technical education system: A linear approach | |
| Wei et al. | Analysis on the Classification and Evaluation System of Talents in Colleges and Universities from the Perspective of AHP | |
| CN116308158B (en) | National asset supervision and management 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: 20180706 |
|
| RJ01 | Rejection of invention patent application after publication |