CN105930353A - Feature subdivision method for call center customer service staff in mass data - Google Patents
Feature subdivision method for call center customer service staff in mass data Download PDFInfo
- Publication number
- CN105930353A CN105930353A CN201610216835.XA CN201610216835A CN105930353A CN 105930353 A CN105930353 A CN 105930353A CN 201610216835 A CN201610216835 A CN 201610216835A CN 105930353 A CN105930353 A CN 105930353A
- Authority
- CN
- China
- Prior art keywords
- data
- cluster
- staff
- clustering
- average
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06Q10/06398—Performance of employee with respect to a job function
Abstract
The invention discloses a feature subdivision method for call center customer service staff in mass data. The feature subdivision method includes: acquiring a preset number of workers; determining a cluster index, and determining data corresponding to the cluster index, of each worker as cluster index data of the worker; performing clustering through a K-Means algorithm according to the cluster index data of each worker so as to obtain a cluster result; and dividing the workers into different classes according to the cluster result to obtain a group result. A call center work management method can automatically adapt worker subdivision work in the mass data, is high in accuracy, and can avoid the situations that classification based on subjective experience does not confirm with actual task performance of the workers.
Description
Technical field
The present invention relates to call center management technical field, more particularly, it relates under a kind of mass data
The feature divided method of call center contact staff.
Background technology
In the management to the operator of call center, generally by manager or teams and groups' pipe of call center
Reason person's subjectivity determines to be broadly divided into operator a few class.But, in view of call center's seat personnel quantity
Reach scale up to ten thousand, this will management level to the subjective understanding of some operator and its actual work performance
Difference is relatively big, thus well can not remove arranged rational related work according to the ability to work of operator.
Accordingly, it would be desirable to a kind of technology can help manager preferably to recognize the general performance of seat personnel of its subordinate,
Preferably carry out such as run enhancing efficiency, attend a banquet arrange an order according to class and grade, the work of the related management such as skills training.
Summary of the invention
It is an object of the invention to provide the staff's feature segmentation side of call center under a kind of mass data
Method, to solve present in prior art the cognitive Bias problem to staff, assists management personnel to enter
Row aid decision.
To achieve these goals, the present invention provides following technical scheme:
A kind of management method of the call center's staff's feature segmentation under mass data, including:
Obtain the KPI achievement data that the staff of the some teams and groups of M row S row examined within some years；
For the pretreatment of described KPI achievement data, described pretreatment mode include missing values supplement,
Replacement and last the process by data normalization of exceptional value obtain clustering target data；
The clustering target data separate distributed K-Means algorithm that will have processed, carries out client characteristics and hives off,
Obtain the segmentation result of client.
Preferably, in described acquisition each described some years, staff examines number corresponding to KPI index
According to, including: percent of call completed, incoming call project occupation rate, exhalation project work efficiency, average handling time,
Average queuing time, rate of solving the problem once and for all, average cost per contact, the rate of attendance, business evaluating achievement
And satisfaction rate；
Wherein, described percent of call completed refers to the connection of the whole level service unit of IVR (Interactive Voice Response)
Measure the ratio of the connection amount sum with manual position and the calling total amount entering call center；
In described incoming call project occupation rate refers to certain section of timing statistics, seat person process take on the telephone more total time
The long ratio with actual log system duration；
In described exhalation project work efficiency refers to certain section of timing statistics, total handling duration and login system duration
Ratio；
Described average handling time: once get in touch with required average time；
Described average queuing time refers in certain section of timing statistics, and caller waits artificial seat after shortlisting
The average waiting duration of the wait that seat is answered；
The described rate of solving the problem once and for all is in certain section of timing statistics, it is not necessary to client dials in calling again
Telephone redialing or switching are accounted for seat person with regard to soluble phone amount and pick up electricity also without seat person by center
The percentage ratio of words total amount；
The described rate of attendance refers in certain statistical time range, and the actual number turned out for work of certain teams and groups is turned out for work with plan
The percentage rate of number；
In described average cost per contact refers to certain section of timing statistics, the full payment of call center divided by phone at
Reason amount；
Described business evaluating achievement refers to that call center's seat is to the Grasping level of professional knowledge and described satisfaction
Degree refers to the service satisfactory degree that call center is provided by client.
Preferably, supplementing of the described missing values that above-mentioned KPI performance assessment criteria data are carried out pretreatment, bag
Include:
Determine total observation data S row, the KPI that the staff of corresponding some teams and groups examined within some years
Index, these indexs are cluster input attribute field；
Data, its attribute field X is observed if there is L rowi(i≤M) has missing values or all inputs all
Record or the observation field with missing values do not have the data corresponding with ATTRIBUTE INDEX, it is determined that this L
Row data need to carry out missing values process；
Determine that residue S-L row data, for reference to data, add up every in these observation data M attribute fields
Individual attribute XiThe value that (i≤M) occurrence number is most, i.e. mode；
Missing data in above-mentioned L row M attribute field of observation is filled with, is replaced with mode,
Complete corresponding attribute XiThe missing values of (i≤M) processes.
Preferably, the replacement of the described abnormal data that above-mentioned KPI performance assessment criteria data are carried out pretreatment,
Including:
Raw sample data S row, each attribute XiThe average of (i≤M) isStandard deviation is δiIf existingThen this observation is an abnormity point.
For abnormal data, ifComplete the replacement of abnormal data.
Preferably, described above-mentioned KPI performance assessment criteria data are normalized, including:
ATTRIBUTE INDEX is carried out standardization processing, can be allowed to by ATTRIBUTE INDEX data bi-directional scaling
Enter a little specific interval；By min-max method for normalizing, utilize following functional transformation formula:
Xmax,XminMaximum and minima, the X of attribute field is concentrated for original samplei,YiIt is respectively input
Value before and after samples normalization；The numerical value x of each clustering target data is mapped to the interval [0,1] of correspondence,
To realize the unitized of different clustering target, obtain the clustering target data with unified metric unit.
Preferably, the described parallel K-Means of clustering target data separate according to each described staff
Algorithm clusters, and obtains cluster result, including:
Initial phase: first determine cluster number k, concentrate from initial data at random and choose k data conduct
Initial classes center, is divided into several data blocks simultaneously by raw data set；Determine class center at random, such as
Original have 10 points, determines that wherein K=2 point is as class center at random.
The MAP stage: each data block (comprising many data) correspondence is distributed to a map function,
For this data block, calculate wherein every data generic；
The REDUCE stage: this stage obtains complete cluster result, specifically by each piece of cluster result of merger
It is to calculate to belong to the average of same category data as Xin Lei center, and as next iteration
Input, until algorithmic statement；Xin Lei center is to be obtained by the mean value computation of same category data.Ratio
As comprised 10 points under a certain class, calculate the average of these 10 points, obtain class central point；
The results verification stage: according to described cluster result, described staff is divided into different classes, obtains spy
Levy segmentation result；
The result application stage: determine the pipe of such staff according to the feature of the staff of each class
Reason mode.
Preferably, in the described results verification stage, need to cluster according to the cluster result of described staff
The determination of optimum number, including:
Initial range is chosen: preset certain cluster number scope [k1,k2], and cluster number is that this is interval
Interior integer；
Inter-object distance calculates: to [k1,k2Any k in], calculates under current cluster result in same class
Distance isWherein xiFor clustering the data point in j,For clustering the apoplexy due to endogenous wind of j
Heart point；
Between class distance calculates: to [k1,k2Any k in], calculates between classes different under current cluster result
Distance beWhereinFor clustering the data point in i,For cluster j's
Class central point；
Optimal cluster number determines: set up Cluster Validity FunctionCarry out cluster result evaluation,
When L (k) is minimum, correspondence obtains optimal cluster number k, thus avoids clusters number and exist on choosing
Subjectivity.
The manager of the call center's staff's feature segmentation under a kind of mass data that the present invention provides
Method, including: obtain the KPI index that the staff of the some teams and groups of M row S row examined within some years
Data；Carrying out the pretreatment of above-mentioned data, including the supplementing of missing values, the replacement of exceptional value, data are returned
One change process etc.；Utilize distributed K-Means algorithm, carry out client characteristics and hive off, obtain the thin of client
Divide result；
Compared with prior art, the application is not simply to set index weights and carries out KPI marking examination,
But it is entered according to the examination KPI achievement data of each staff by parallel K-Means algorithm
Row hives off, and obtains the general cognitive about seat personnel task performance.And K-Means algorithm is pin parallel
Cluster under mass data, the iteration of calculating process can be completed more quickly, can be the completeest
Become correlated characteristic segmentation work, therefore, work people in call center's under a kind of mass data disclosed in the present application
The management method of member's feature segmentation, analytical performance is higher, it is to avoid manager uprushes because of crew numbers
And cause the performance cognition of the real work to staff to occur that the situation of deviation occurs.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to reality
Execute the required accompanying drawing used in example or description of the prior art to be briefly described, it should be apparent that below,
Accompanying drawing in description is only embodiments of the invention, for those of ordinary skill in the art, not
On the premise of paying creative work, it is also possible to obtain other accompanying drawing according to the accompanying drawing provided.
Staff's feature segmentation in call center's under a kind of mass data that Fig. 1 provides for the embodiment of the present invention
The flow chart of management method；
The stream of the Missing Data Filling of a kind of examination KPI data pretreatment that Fig. 2 provides for the embodiment of the present invention
Cheng Tu；
The replacement of the exceptional value of a kind of examination KPI data pretreatment that Fig. 3 provides for the embodiment of the present invention
Flow chart；
The stream of the data normalization of a kind of examination KPI data pretreatment that Fig. 4 provides for the embodiment of the present invention
Cheng Tu；
The Parallel K-Means Clustering Algorithm in Web that utilizes that Fig. 5 provides for the embodiment of the present invention is operated personnel characteristics
Segmentation, obtains the flow chart of cluster result；
Fig. 6 determines the optimal stream clustering number for what the embodiment of the present invention provided according to Cluster Validity Function
Cheng Tu；
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out
Clearly and completely describe, it is clear that described embodiment is only a part of embodiment of the present invention, and
It is not all, of embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art are not doing
Go out the every other embodiment obtained under creative work premise, broadly fall into the scope of protection of the invention.
Refer to Fig. 1, it illustrates the flow process of a kind of staff's management method that the embodiment of the present invention provides
Figure, may include that
S11: obtain the KPI index number that the staff of the some teams and groups of M row S row examined within some years
According to；Including percent of call completed, incoming call project occupation rate, exhalation project work efficiency, average handling time, put down
All queuing time, rate of solving the problem once and for all, average cost per contact, rate of attendance business evaluating achievements and full
Meaning rate；
Table 1KPI index and implication thereof
Choose the some employees in call center job performance data in specific duration, carry out feature clustering.
Concrete sample data example such as following table:
The sample data of table 2KPI index
Predetermined amount can be determined according to actual needs, is not specifically limited at this.
S12: carry out the pretreatment of above-mentioned data, including missing values supplement, the replacement of exceptional value, data
Normalized obtains clustering target data；
Wherein, clustering target is the index in terms of the work performance of staff, and its concrete meaning is according to work
The position making personnel is different and otherwise varied.It addition, the quantity of clustering target can be 1, it is also possible to
For multiple, it is determined according to actual needs, is not specifically limited at this.
S13: utilize distributed K-Means algorithm, carries out client characteristics and hives off, and obtains the segmentation knot of client
Really；
Wherein, the object in set is divided into the data analysing method of several classes by cluster exactly, and it makes
Between the object of each apoplexy due to endogenous wind the most similar, the object of different apoplexy due to endogenous wind is the most different, the application
In object be staff.
The application by parallel K-means clustering algorithm according to the clustering target data pair of each staff
It is classified, and staff is classified by the cluster result utilizing clustering algorithm to obtain, it is achieved each
Staff's maximizing in class similar.Therefore, a kind of staff manager disclosed in the present application
Method is not it can be avoided that staff's cognition of causing because of the subjective impression of upper-level leader meets practical situation and sends out
Raw.
It addition, Parallel K-Means Clustering Algorithm in Web is applied to the classification for staff by the present invention, have
It is beneficial to realize the lean management of staff, it is simple to the leader of call center understands the operation at center in time
The problem existed during efficiency and service, and follow the tracks of action plan and performance result by correction strategy,
Thus ensure the realization of development strategy.
Refer to Fig. 2, a kind of examination KPI data pretreatment of embodiment of the present invention offer is provided
The flow chart of Missing Data Filling, may include that
S21: acquiring unit.Obtaining total observation data S row, the staff of corresponding some teams and groups is some
The KPI index of examination in year, determines that these indexs are for cluster input attribute field；
S22: detector unit.Data, its attribute field X is observed if there is L rowi(i≤M) has disappearance
Value or all inputs all have the record of missing values or observation field does not have the number corresponding with ATTRIBUTE INDEX
According to, then detect and determine that these L row data need to carry out missing values process；
S23: computing unit.Determine that residue S-L row data, for reference to data, add up these observation data M
Each attribute X in individual attribute fieldiThe value that (i≤M) occurrence number is most, i.e. mode；
S24: fill unit.Missing data in above-mentioned L row M attribute field of observation is filled with,
It is replaced with mode, completes corresponding attribute XiThe missing values of (i≤M) processes.
Refer to Fig. 3, a kind of examination KPI data pretreatment of embodiment of the present invention offer is provided
The flow chart of the replacement of exceptional value, may include that
S31: confirmation unit.Raw sample data S row, each attribute XiThe average of (i≤M) isMark
Quasi-difference is δiIf existingThen confirm that this observation is an abnormity point.
S32: replacement unit.For abnormal data, ifComplete abnormal number
According to replacement.
Refer to Fig. 4, a kind of examination KPI data pretreatment of embodiment of the present invention offer is provided
The flow chart of data normalization；
S41: computing unit.Calculate original sample and concentrate maximum and the minima of each attribute field
Xmax,Xmin；
S42: normalization unit.By min-max method for normalizing, utilize following functional transformation formula:
Wherein Xi,YiIt is respectively the value before and after input samples normalization.Numerical value by each clustering target data
X is mapped to the interval [0,1] of correspondence.
By ATTRIBUTE INDEX is carried out standardization processing, ATTRIBUTE INDEX data bi-directional scaling can be made
Fall into a little specific interval.To realize the unitized of different clustering target, obtain that there is unified degree
The clustering target data of amount unit.
Referring to Fig. 5, it illustrates embodiment of the present invention offer utilizes Parallel K-Means Clustering Algorithm in Web
It is operated personnel characteristics's segmentation, obtains the flow chart of cluster result；
S51: initialization unit.First determine cluster number, determine initial all kinds of centers at random, simultaneously
Raw data set is divided into several data blocks.
S52:MAP unit.Each data subset correspondence is distributed to a map function, for often
Individual data block, calculates each data item generic.
S53:REDUCE unit.Each piece of cluster result of merger obtains complete cluster result, recalculates
Class center is as the input of next iteration, until algorithmic statement.
S54: results verification unit.According to described cluster result, described staff is divided into different classes,
Obtain feature segmentation result.
Illustrate above-mentioned steps.As operator can be divided into following 3 classes:
Cluster 1 (maker post type): the indexs such as its KPI of the operator related to do well, percent of call completed, incoming call item
Mesh occupation rate, exhalation project work efficiency, rate of solving the problem once and for all, the rate of attendance, business evaluating achievement
And satisfaction rate is the highest, but average handling time, average queuing time and average cost per contact are the most relatively low, say
Bright this kind of people's work quality is the highest, and explanation work efficiency is high simultaneously.
Cluster 2 (treating concern type): the indexs such as its KPI of the operator related to do well, percent of call completed, incoming call
Project occupation rate, exhalation project work efficiency, rate of solving the problem once and for all, the rate of attendance, business evaluating become
Achievement and satisfaction rate are minimum, and inefficiency is described.Average handling time, average queuing time and the most single
Exhale cost the highest, illustrate that work quality is minimum.
Cluster 3 (middle types): the percent of call completed of this kind of operator, incoming call project occupation rate, exhalation project work
Make efficiency, rate of solving the problem once and for all, the rate of attendance, business evaluating achievement and satisfaction rate and be in by-level,
Average handling time, average queuing time and average cost per contact are in by-level.
S55: result applying unit.The feature of the staff according to each class determines such staff
Way to manage.
In a kind of staff's management method that above-described embodiment provides, according to cluster result by staff
It is divided into different classes, after obtaining classification results, determines the staff's of each class according to classification results
Feature carries out fine-grained management, promotes the efficiency of operation of call center；
Citing carries out above-mentioned steps explanation.
One is job content and the work that the feature of the staff according to each class determines such staff
Make the time.For " maker post type ", " middle type " and the operator of " treating concern type ", take correspondence
Excitation penal system, is operated the arrangement of content and working time, wherein, when job content and work
Between specifically can include arranging an order according to class and grade the diversity arrangement etc. of duration and services client, such as " problem type " personnel's night shift
Time of arranging an order according to class and grade is longer, the group customer of " maker post type " personnel service's Electric Power Customer Service Center and VIP client.
Thus, according to the feature of different operating personnel, Reasonable adjustment order of classes or grades at school, and corresponding different demand (as
Customer demand etc.), distribute the staff of different features, be conducive to the management for staff, with
And improve the work efficiency of staff.
Two is the work defect that the feature of the staff according to each class determines such staff, and
For the staff of each class, the staff under class of especially " treating concern type ", according to its work
Such staff is giveed training by defect.The work of the staff of each class is determined based on grouping result
Making defect, beneficially manager arranges by a definite date difference according to the work defect difference of the staff of each class
Training, the beneficially raising of staff's level.On this basis, long-acting training mechanism, shape are set up
Become the work climate of autonomic learning.
Refer to Fig. 6, determining most preferably according to Cluster Validity Function of embodiment of the present invention offer is provided
The flow chart of cluster number；
S61: initial range is chosen.Preset certain cluster number scope [k1,k2], and cluster number be this
Integer in interval；
S62: inter-object distance calculates.To [k1,k2Any k in], calculates under current cluster result same
Distance in class isWherein xiFor clustering the data point in j,For cluster j's
Class central point；
S63: between class distance calculates.To [k1,k2Any k in], calculates under current cluster result different
Distance between class isWhereinFor clustering the data point in i,For cluster
The class central point of j；
S64: optimal cluster number determines.Set up Cluster Validity FunctionCarry out cluster knot
Fruit is evaluated, and when L (k) is minimum, correspondence obtains optimal cluster number k, thus avoids clusters number in choosing
Subjectivity present on taking.
Described above to the disclosed embodiments, makes those skilled in the art be capable of or uses this
Bright.Multiple amendment to these embodiments will be apparent from, herein for a person skilled in the art
Defined in General Principle can be real at other without departing from the spirit or scope of the present invention
Execute in example and realize.Therefore, the present invention is not intended to be limited to the embodiments shown herein, and is intended to
Meet the widest scope consistent with principles disclosed herein and features of novelty.
Claims (8)
1. the method for the feature segmentation of call center contact staff under a mass data, it is characterised in that bag
Include:
Obtain the KPI achievement data that the staff of the some teams and groups of M row S row examined within some years；
For the pretreatment of described KPI achievement data, described pretreatment mode includes the supplementary, different of missing values
Replacement and last the process by data normalization of constant value obtain clustering target data；
The clustering target data separate distributed K-Means algorithm that will have processed, carries out client characteristics and hives off,
Obtain the segmentation result of client.
Method the most according to claim 1, it is characterised in that described acquisition each described some years
Interior staff examines the data that KPI index is corresponding, including: percent of call completed, incoming call project occupation rate, exhalation
Project work efficiency, average handling time, average queuing time, rate of solving the problem once and for all, average individual calling
Cost, the rate of attendance, business evaluating achievement and satisfaction rate；
Wherein, described percent of call completed refers to the connection amount of IVR level at end service unit and the connection amount sum of manual position
Ratio with the calling total amount entering call center；In described incoming call project occupation rate refers to certain section of timing statistics,
Seat person processes the ratio of total duration and the actual log system duration taken on the telephone more；Described exhalation project work
In efficiency refers to certain section of timing statistics, total handling duration and the ratio of login system duration.During described average treatment
Between: once get in touch with required average time；Described average queuing time refers in certain section of timing statistics, calling
Person waits the average waiting duration of the wait of manual position answer after shortlisting；Described solve the problem once and for all
Rate is in certain section of timing statistics, it is not necessary to client dials in call center also without seat person again by phone
Clawback or switching account for seat person with regard to soluble phone amount and pick up the percentage ratio of phone total amount；The described rate of attendance
Refer in certain statistical time range, the percentage rate of the number that the actual number turned out for work of certain teams and groups is turned out for work with plan；
In described average cost per contact refers to certain section of timing statistics, the full payment of call center is divided by phone treating capacity；
Described business evaluating achievement refers to call center's seat Grasping level to professional knowledge；Described satisfaction refers to client
The service satisfactory degree that call center is provided.
3. want the method described in 1 according to right, it is characterised in that missing values is supplemented, including:
Determine total observation data S row, the KPI that the staff of corresponding some teams and groups examined within some years
Index, these indexs are cluster input attribute field；
Data, its attribute field X is observed if there is L rowi(i≤M) has missing values or all inputs all have
The record or the observation field that have missing values do not have the data corresponding with ATTRIBUTE INDEX, it is determined that this L line number
Missing values process is carried out according to needs；
Determine that residue S-L row data, for reference to data, add up each in these observation data M attribute fields
Attribute XiThe value that (i≤M) occurrence number is most, i.e. mode；
Missing data in above-mentioned L row M attribute field of observation is filled with, is replaced with mode,
Complete corresponding attribute XiThe missing values of (i≤M) processes.
Method the most according to claim 1, it is characterised in that for the replacement of abnormal data, including:
Raw sample data S row, each attribute XiThe average of (i≤M) isStandard deviation is δiIf existingThen this observation is an abnormity point；
For abnormal data, ifComplete the replacement of abnormal data.
5. according to the arbitrary described method of Claims 1-4, it is characterised in that for above-mentioned KPI index
Data are all normalized, including:
ATTRIBUTE INDEX is carried out standardization processing, can be allowed to fall into by ATTRIBUTE INDEX data bi-directional scaling
One little specific interval, by min-max method for normalizing, utilizes following functional transformation formula:
Xmax,XminMaximum and minima, the X of attribute field is concentrated for original samplei,YiRespectively input sample
Value before and after this normalization.The numerical value x of each clustering target data is mapped to the interval [0,1] of correspondence, with
Realize the unitized of different clustering target, obtain the clustering target data with unified metric unit.
Method the most according to claim 5, it is characterised in that described according to each described staff
Clustering target data separate parallel K-Means algorithm cluster, obtain cluster result, including:
Initial phase: first determine cluster number, determines initial all kinds of centers, at random simultaneously by original number
It is divided into several data blocks according to collection；
The MAP stage: each data block correspondence is distributed to a map function, for this data block, meter
Calculating wherein every data generic, above-mentioned data base comprises many data；
The REDUCE stage: this stage obtains complete cluster result, specifically by each piece of cluster result of merger
Calculate and belong to the average of same category data as Xin Lei center, and defeated as next iteration
Enter, until algorithmic statement；
The results verification stage: according to described cluster result, described staff is divided into different classes, obtains feature
Segmentation result；
The result application stage: determine the management of such staff according to the feature of the staff of each class
Mode.
Method the most according to claim 6, it is characterised in that in the described results verification stage, need basis
The cluster result of described staff carries out clustering the determination of optimum number, including:
Preset certain cluster number scope [k1,k2], and cluster number is the integer in this interval；
To [k1,k2Any k in], calculates the distance in same class under current cluster result；
To [k1,k2Any k in], calculates the distance between classes different under current cluster result；
Determine Cluster Validity Function, carry out cluster result evaluation, obtain optimal cluster number, thus avoid
Clusters number subjectivity present on choose.
Method the most according to claim 7, it is characterised in that in the described results verification stage, need basis
The cluster result of described staff carries out clustering the determination of optimum number, including:
Initial range is chosen: preset certain cluster number scope [k1,k2], and cluster is in number is this interval
Integer；
Inter-object distance calculates: to [k1,k2Any k in], calculate under current cluster result in same class away from
From forWherein xiFor clustering the data point in j,For clustering the class central point of j；
Between class distance calculates: to [k1,k2Any k in], calculates between classes different under current cluster result
Distance isWhereinFor clustering the data point in i,For clustering the apoplexy due to endogenous wind of j
Heart point；
Optimal cluster number determines: set up Cluster Validity FunctionCarry out cluster result evaluation,
When L (k) is minimum, correspondence obtains optimal cluster number k, thus avoids clusters number present on choose
Subjectivity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610216835.XA CN105930353A (en) | 2016-04-08 | 2016-04-08 | Feature subdivision method for call center customer service staff in mass data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610216835.XA CN105930353A (en) | 2016-04-08 | 2016-04-08 | Feature subdivision method for call center customer service staff in mass data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105930353A true CN105930353A (en) | 2016-09-07 |
Family
ID=56840532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610216835.XA Pending CN105930353A (en) | 2016-04-08 | 2016-04-08 | Feature subdivision method for call center customer service staff in mass data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105930353A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529168A (en) * | 2016-11-08 | 2017-03-22 | 无锡市妇幼保健院 | Gynecological disease intelligent diagnosis oriented data preprocessing technology |
| CN106875224A (en) * | 2017-02-22 | 2017-06-20 | 山东浪潮云服务信息科技有限公司 | Tourism big data customer segmentation method based on MapReduce |
| CN107274066A (en) * | 2017-05-19 | 2017-10-20 | 浙江大学 | A kind of shared traffic Customer Value Analysis method based on LRFMD models |
| CN108322615A (en) * | 2018-02-01 | 2018-07-24 | 福建工程学院 | A kind of more technical ability calling customer service system method for optimizing configuration |
| CN108573266A (en) * | 2017-03-10 | 2018-09-25 | 中国移动通信集团河北有限公司 | The method and apparatus for extracting common trait |
| CN108664000A (en) * | 2018-03-26 | 2018-10-16 | 中南大学 | A kind of alumina producing evaporation process steady state detecting method for use and system |
| CN109214153A (en) * | 2017-07-05 | 2019-01-15 | 百度在线网络技术（北京）有限公司 | information generating method and device |
| CN109816034A (en) * | 2019-01-31 | 2019-05-28 | 清华大学 | Signal characteristic combines choosing method, device, computer equipment and storage medium |
| CN110147905A (en) * | 2019-05-08 | 2019-08-20 | 联想(北京)有限公司 | Information processing method, device, system and storage medium |
| CN110472054A (en) * | 2019-08-15 | 2019-11-19 | 北京爱数智慧科技有限公司 | A kind of data processing method and device |
| CN112184051A (en) * | 2020-10-13 | 2021-01-05 | 中国工程物理研究院计算机应用研究所 | Employee work investigation method based on social network analysis technology |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040210400A1 (en) * | 2003-01-27 | 2004-10-21 | Perlegen Sciences, Inc. | Analysis methods for individual genotyping |
| CN104850629A (en) * | 2015-05-21 | 2015-08-19 | 杭州天宽科技有限公司 | Analysis method of massive intelligent electricity-consumption data based on improved k-means algorithm |
| CN105117869A (en) * | 2015-10-16 | 2015-12-02 | 北京中电普华信息技术有限公司 | Worker management method |
-
2016
- 2016-04-08 CN CN201610216835.XA patent/CN105930353A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040210400A1 (en) * | 2003-01-27 | 2004-10-21 | Perlegen Sciences, Inc. | Analysis methods for individual genotyping |
| CN104850629A (en) * | 2015-05-21 | 2015-08-19 | 杭州天宽科技有限公司 | Analysis method of massive intelligent electricity-consumption data based on improved k-means algorithm |
| CN105117869A (en) * | 2015-10-16 | 2015-12-02 | 北京中电普华信息技术有限公司 | Worker management method |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529168A (en) * | 2016-11-08 | 2017-03-22 | 无锡市妇幼保健院 | Gynecological disease intelligent diagnosis oriented data preprocessing technology |
| CN106875224A (en) * | 2017-02-22 | 2017-06-20 | 山东浪潮云服务信息科技有限公司 | Tourism big data customer segmentation method based on MapReduce |
| CN108573266A (en) * | 2017-03-10 | 2018-09-25 | 中国移动通信集团河北有限公司 | The method and apparatus for extracting common trait |
| CN107274066A (en) * | 2017-05-19 | 2017-10-20 | 浙江大学 | A kind of shared traffic Customer Value Analysis method based on LRFMD models |
| CN109214153A (en) * | 2017-07-05 | 2019-01-15 | 百度在线网络技术（北京）有限公司 | information generating method and device |
| CN108322615A (en) * | 2018-02-01 | 2018-07-24 | 福建工程学院 | A kind of more technical ability calling customer service system method for optimizing configuration |
| CN108322615B (en) * | 2018-02-01 | 2020-07-28 | 福建工程学院 | Configuration optimization method for multi-skill calling customer service system |
| CN108664000A (en) * | 2018-03-26 | 2018-10-16 | 中南大学 | A kind of alumina producing evaporation process steady state detecting method for use and system |
| CN109816034A (en) * | 2019-01-31 | 2019-05-28 | 清华大学 | Signal characteristic combines choosing method, device, computer equipment and storage medium |
| CN110147905A (en) * | 2019-05-08 | 2019-08-20 | 联想(北京)有限公司 | Information processing method, device, system and storage medium |
| CN110472054A (en) * | 2019-08-15 | 2019-11-19 | 北京爱数智慧科技有限公司 | A kind of data processing method and device |
| CN112184051A (en) * | 2020-10-13 | 2021-01-05 | 中国工程物理研究院计算机应用研究所 | Employee work investigation method based on social network analysis technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105930353A (en) | Feature subdivision method for call center customer service staff in mass data | |
| CN102158879B (en) | Essential factor lost score data processing method and equipment | |
| CN101951617B (en) | Mobile network communication quality evaluation method based on analytic hierarchy process | |
| CN104680428B (en) | A kind of construction method of power network CSAT model | |
| CN110110881B (en) | Power customer demand prediction analysis method and system | |
| CN105096058A (en) | Data processing method and device for customer service staff scoring system | |
| CN106228286A (en) | A kind of data analysing method for the assessment of artificial customer service work quality | |
| CN101562832B (en) | Method and system for calculating key quality index used in communication test | |
| CN101777203B (en) | Centralized management system for banking service | |
| CN105205341B (en) | A kind of power matching network transformation demand model construction method based on client's demand | |
| CN107392479A (en) | The power customer power failure susceptibility scorecard implementation of logic-based regression model | |
| CN102163310A (en) | Information pushing method and device based on credit rating of user | |
| CN107403285A (en) | Operator's marketing channel Valuation Method based on optimal segmentation | |
| CN108495329A (en) | A kind of evaluation method and device of base station reliability | |
| CN106771862B (en) | A kind of acquisition abnormity fault point polymerization that grid is combined with space length | |
| CN106530139A (en) | Method for calculating the index parameter of grid investment analysis model | |
| CN109064023A (en) | A kind of method and apparatus of manpower potency management system | |
| CN108268477A (en) | A kind of customer relationship based on ticket determines method and apparatus | |
| CN103902798B (en) | Data preprocessing method | |
| CN110570097A (en) | business personnel risk identification method and device based on big data and storage medium | |
| CN116307841A (en) | Construction method of college evaluation system integrating multiple indexes | |
| CN108924369A (en) | NPS assessment method for running, systems and management platform based on telephone poll | |
| CN104217088B (en) | The optimization method and system of operator's mobile service resource | |
| Deliktas et al. | Economic growth and input use efficiency in low, upper-middle and high incomed countries (1991-2011): a data envelopment analysis | |
| CN114186781A (en) | Impact load identification method and system based on sampling interval load curve |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 710065 Shaanxi hi tech Zone, Xi'an science and technology zone, No. 68 Xi'an Software Park, Han Yun Ge C501 Applicant after: Merrill Lynch data technology Limited by Share Ltd Address before: 710065 Shaanxi hi tech Zone, Xi'an science and technology zone, No. 68 Xi'an Software Park, Han Yun Ge C501 Applicant before: XI'AN MERIT DATA TECHNOLOGY CO., LTD. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160907 |