CN106303113A - A kind of shunt method and equipment - Google Patents

Abstract

This application discloses a kind of shunt method and equipment, including: receive user's request；The content asked based on described user judges the type of customer problem；Type based on described customer problem and shunting model, carry out shunting prediction to described user request；Result based on described shunting prediction, branches to process accordingly object by described user request；Achieving the intellectuality of branching process with this, shunting is quick and accurate, improves the treatment effeciency processing object, saves human resources, better user experience, enhance user's sense of trust to company.

Description

A kind of shunt method and equipment

Technical field

The invention relates to networking technology area, particularly to a kind of shunt method.The application implements Example also relates to a kind of shunting device.

Background technology

In prior art, hotline service need to put into the outsourcing service situation that manual observations is instant, according to the observation As a result, artificially regulate shunt ratio, to reach to maximally utilize the purpose of outsourcing resource, such as, if User's demand queuing amount of current outsourcing company A is bigger, and outsourcing company B can process more user Demand, then with regard to the artificial shunt ratio tuning up outsourcing company B, so that outsourcing company B can obtain more Demand flow.When the service of all outsourcing companies is all in saturation, just it is based entirely on related work The experience of personnel regulates shunt ratio.First substantial amounts of human resources, and people are consumed Failure risk on line can be added for operation；Secondly, the service quality of outsourcing company has dividing of height, as Really outsourcing company A and outsourcing company B can process user demand S, but company A is than the Service Quality of B company Amount height, then obviously demand is distributed to company A and can improve Consumer's Experience, strengthen user's trust to customer service Sense, but, manual adjustment shunt ratio is at random demand S to be assigned to A and B, it is impossible to reach this effect Really, it addition, when the service of outsourcing company is all in saturation, manual adjustment shunt ratio, easily Make empiricism mistake, because the ability to predict of people can change according to time, the change of environment, so Artificial Control shunt ratio, it is impossible to make the undertaking amount of outsourcing company maximize.

Currently also there is simple automatic shunt, be the business throughput (unit by calculating outsourcing company The user's demand number processed in time), in this, as the foundation of outsourcing shunting, user can be told by this scheme Ask and distribute to the outsourcing company that business throughput is maximum.But, there is an obvious defect, business in this The computing formula of handling capacity is:

(the user's demand number processed in the past period)/(time interval),

Consider that, in the case of queuing up, the number of users processed in the past period is the most, then just have Corresponding new demand takies customer service resource, will liberate these customer service resources, take longer for, so leading Cause outsourcing company and process the ability decline of current business demand the soonest.Therefore, this shunt method is not suitable for Queuing situation.

To sum up, in the face of business is more and more various, the situation that customer issue becomes increasingly complex, existing shunting Scheme is the most intelligent, it is considered to factor single, cause shunt result imperfection, to offering customers service Inefficient.

Summary of the invention

The embodiment of the present application provides a kind of shunt method and equipment, in this application, it is achieved that quickly And shunt accurately, improve the treatment effeciency processing object, branching process is intelligent, saves a large amount of Human resources, it is to avoid the error of manual operation, reduces failure risk on line, improves Consumer's Experience, Enhance user's sense of trust to company.

To this end, the embodiment of the present application provides a kind of shunt method, including:

Receive user's request；

The content asked based on described user judges the type of customer problem；

Type based on described customer problem and shunting model, carry out shunting prediction to described user request；

Result based on described shunting prediction, branches to process accordingly object by described user request.

Preferably, the described content asked based on described user judges the type of customer problem, specifically includes:

Receive described user request, it is thus achieved that the customer problem of speech form describes；

The customer problem of described speech form is described the customer problem description changing into textual form；

Customer problem description to described textual form is modified；

Identify that the customer problem of revised textual form describes, determine the type of customer problem.

Preferably, described type based on described customer problem and shunting model, described user is asked into Row shunting prediction, specifically includes:

Type based on described customer problem, determines alternative process object；

Described alternative process object is carried out prioritization；

Determine that described shunting predicts the outcome based on described prioritization result；

Preferably, described described alternative process object is carried out prioritization, specifically includes:

Based on the priority parameters calculated, described alternative process object is carried out prioritization；

Wherein, described priority parameters is

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}lnF+\mathrm{\δ}ln(P-T),(P>T);$

Wherein, processing user in S is the described process object one section of Preset Time after current time please The quantized value of the ability asked；C be current time before one section of Preset Time in described process object receive use The quantity of family request；R is the quantity of the pending user request of described process object；N is described process The quantity of processing unit in object；F is the quantity of idle processing unit in described process object；T is described Process object and the most receive the quantity of user's request；P is that described process object receives estimating of user's request Amount；α, β, γ, δ are parameter.

Preferably, the determination mode of described parameter, particularly as follows:

Determine by presetting；And/or,

Determined by the process of supervised learning；And/or,

Determined by the process of adaptive learning.

The embodiment of the present application also discloses a kind of shunting device, including:

Receiver module, is used for receiving user's request；

Judge module, the content for asking based on described user judges the type of customer problem；

Prediction module, for type based on described customer problem and shunting model, asks described user Carry out shunting prediction；

Diverter module, for result based on described shunting prediction, branches to corresponding by described user request Process object.

Preferably, described judge module, specifically include:

Receive described user request, it is thus achieved that the customer problem of speech form describes；

The customer problem of described speech form is described the customer problem description changing into textual form；

Customer problem description to described textual form is modified；

Identify that the customer problem of revised textual form describes, determine the type of customer problem.

Preferably, described prediction module, specifically include:

Type based on described customer problem, determines alternative process object；

Described alternative process object is carried out prioritization；

Determine that described shunting predicts the outcome based on described prioritization result；

Preferably, described prediction module carries out prioritization to described alternative process object, specifically includes:

Based on the priority parameters calculated, described alternative process object is carried out prioritization；

Wherein, described priority parameters is

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T);$

Wherein, processing user in S is the described process object one section of Preset Time after current time please The quantized value of the ability asked；C be current time before one section of Preset Time in described process object receive use The quantity of family request；R is the quantity of the pending user request of described process object；N is described process The quantity of processing unit in object；F is the quantity of idle processing unit in described process object；T is described Process object and the most receive the quantity of user's request；P is that described process object receives estimating of user's request Amount；α, β, γ, δ are parameter.

Preferably, the determination mode of described parameter, particularly as follows:

Determine by presetting；And/or,

Determined by the process of supervised learning；And/or,

Determined by the process of adaptive learning.

Compared with prior art, by receiving user's request in the application；In asking based on described user Hold the type judging customer problem；Type based on described customer problem and shunting model, to described user Request carries out shunting prediction；Result based on described shunting prediction, branches to corresponding by described user request Process object, with this by simple step, it is achieved that quickly and accurately shunt, improve process The treatment effeciency of object, branching process is more intelligent, save a large amount of human resources, it is to avoid artificial The error of operation, reduces failure risk on line.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of a kind of shunt method in application the embodiment of the present application；

Fig. 2 is the schematic flow sheet of the shunt method under the concrete scene that the embodiment of the present application proposes；

Fig. 3 is the sequential chart of the shunt method under the concrete scene that the embodiment of the present application proposes；

Fig. 4 is the structural representation of a kind of shunting device that the embodiment of the present application proposes.

Detailed description of the invention

For the above-mentioned problems in the prior art, the embodiment of the present application provide a kind of shunt method and Equipment, in order to overcome defect of the prior art.

Below in conjunction with the accompanying drawings the embodiment of the present invention is described in detail.

The embodiment of the present application one provides a kind of shunt method and equipment, as it is shown in figure 1, include following step Rapid:

Step S101, reception user's request；

User initiates request to system, and such as user calls, or uses other can carry out with system Mutual mode, system is set up with user after being connected, and can be provided about it by information prompting user The statement of problem, obtains the content of user's request with this, and such as, system is default by playing one section How voice provides its problem to instruction manual, and user can correctly provide its problem according to the step in explanation Statement.

Step S102, the content asked based on described user judge the type of customer problem；

User that system receives request, and user to provide to system be that the problem of speech form is retouched State, then system is first by ARS (Automatic Speech Recognition, automatic speech recognition skill Art) or other similar techniques, the customer problem of described speech form is described and changes into textual form Customer problem describes, and then the customer problem to described textual form describes and is modified, and finally identifies and repaiies The customer problem of the textual form after just describes, and determines the type of customer problem.

Wherein, the customer problem that the customer problem description of described speech form changes into textual form describes Time, the pause of recordable user, when the customer problem to described textual form describes and is modified afterwards, Can with user pause and " " etc. word for interval make pauses in reading unpunctuated ancient writings, customer problem is divided into some unit, The conventional character library of the relevant business information by pre-building combines the existing basic number for text error correction According to storehouse, use the mode of comparison, or take the similar mode searched for generally, first can search text In lard speech with literary allusions this time common mistake either with or without voice, if found, be replaced with correct words, Secondly threshold value can be set, if certain sentence unit and the everyday expressions similarity relating to business information surpass Cross threshold value to be then replaced.

Finally identifying that revised problem describes, can extract the keyword in statement, keyword is generally Verb in text and noun, the various combination of some key words forms different problem typess, such as " money Gold ", the customer problem that combines representative with " robber ", " loss ", " lacking " etc. such as " money " be exactly typically fund Stolen.So, by set up keyword with and combinations thereof with the data base of problem types corresponding relation, i.e. Can determine that customer problem type.

Step S103, type based on described customer problem and shunting model, carried out described user request Shunting prediction；

It is primarily based on the type of the customer problem determined in step S102, determines alternative process object, wherein, The described alternative process determined, to liking this type of customer problem of special disposal, can be to provide customer service Company in technical ability group, such as have several companies, several companies have again several technical ability groups, If that customer problem type to be fund stolen, then all companies solve the skill of the stolen problem of fund Group can be determined as alternative process object.

Then, described alternative process object being carried out prioritization, sequence is excellent based on calculate First level parameter, priority parameters is X,

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T);$

Wherein, processing user in S is the described process object one section of Preset Time after current time please The quantized value of the ability asked；C be current time before one section of Preset Time in described process object receive use The quantity of family request；R is the quantity of the pending user request of described process object；N is described process The quantity of processing unit in object；F is the quantity of idle processing unit in described process object；T is described Process object and the most receive the quantity of user's request；P is that described process object receives estimating of user's request Amount；α, β, γ, δ are parameter, represent the importance degree of each data when calculating priority parameters X, If the business rules that should meet in view of described process object, then δ is not 0, and wherein, business rules refers to, Process the workload that needed of object or other rules pre-established, such as customer service company Treating capacity need to reach to estimate in advance day in one day, can exceed 15% estimated；

Wherein, if there is the process object that X value is identical, can be with the identical place of X value described in random arrangement Sequencing between reason object, it is also possible to obtain the process evaluation in Preset Time of the described process object Mark, the process object priority sequence processing evaluation score high is forward.

Determining that described shunting predicts the outcome based on described prioritization result, generally shunting predicts the outcome and is This user request is branched to the process object of prioritization first.

Wherein, the determination mode of described parameter has three kinds:

The first, determine by presetting；

δ-value in formula can determine by presetting, specifically can be according to the degree of deviation business rules, such as The workload that process object currently completes is much smaller than the workload needed, the most predeterminable bigger δ-value.

The second, is determined by the process of supervised learning；

Supervised learning refers to the parameter utilizing the sample of one group of known class to adjust grader so that it is reach institute Requiring the process of performance, also referred to as supervised training or have teacher learning, supervised learning is the training from labelling Data infer the machine learning task of a function.Training data includes a set of training example.In supervision In study, each example is (also referred to as to be supervised letter by an input object and a desired output valve Number) composition.Supervised learning algorithm is to analyze this training data, and produces a function inferred.

In this application, training data has a F, C, R, N, Label (1), Label (2), Label (3), Label (4), wherein, it is the artificial scoring that i-th processes object that Label (i) represents shunting result.

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T)$

$L={\mathrm{\Σ}}_{i=1}^{n}Score(fi,ci,ri,ni)$ (object function)

Score (fi, ci, ri, ni) is the score of the i-th data shunting result.

Wherein, the target of training is to make target function value L maximum.

The third, determined by the process of adaptive learning.

Adaptive learning typically refers to environment, example or the field domain providing corresponding study in study, logical Cross learner and sum up from finding in study, ultimately form theory can the autonomous side of solving learning of problem Formula.Adaptive learning in this programme refers to: can regulate self model ginseng according to the quality of objective result Number, makes the result of decision more excellent.

In this application, the initial value of α, β, γ can be set to 1, step-length c is set to 0.01, with The value of α, β, γ carries out by machine+,-operation, training natural law is set to U, the user of statistics every day is flat In equal waiting time and process object, the average treatment amount of each processing unit is as evaluation criteria, if User's average waiting time length ratio was short for upper one day, and the average treatment amount of each processing unit is more than upper one day, Then parameter is evolved successfully, when training natural law is equal to U, stops study, and U can be arranged as the case may be.

Determine that mode can be one therein, or according to practical situation, select in three of the above mode Two or three combination, to reach optimum shunting effect.

Step S104, result based on described shunting prediction, branch to locate accordingly by described user request Reason object；

The process object determined in result with reference to described shunting prediction, branches to most suitable by user's request Process object handles, can be specifically by a platform, carry out in the way of demand group is single, as will At the corresponding technical ability group of the company being forwarded to determine of the phone of user.

Below in order to illustrate the application further, the embodiment of the present application two proposes a kind of concrete field Shunt method under scape, as in figure 2 it is shown, include:

Step S201, user describe the problem of oneself by phone, and system obtains the problem of speech form and retouches State；

Customer service is generally divided into Self-Service, online service and hotline service three kinds, listening user phone After, can determine that user there is a problem when being and use which kind of business by the form of voice menu, Yi Jiyong Which kind of family needs service, and user can select by the way of button, needs if now determining user Want hotline service, then prompting user can be illustrated starting at its problem, and informs the instruction that user profile is complete. Such as, user puts through customer service hot line, system prompt ", please by 1, Taobao's business is please by 2 for Alipay business ... ", User select corresponding after, system prompt " please by 1, hotline service is please by 2 in Self-Service ... " if User have selected hotline service, then system prompts user can be illustrated starting at its problem, has described its problem After with " # " bond bundle, the most also can add the step allowing user input its identity information, after Continuous processing procedure can be more targeted, as it is shown on figure 3, this process can pass through CSIVR (interactive voice Platform is supported in response) realize, the effect of CSIVR is to provide for IVR (interactive voice answering system) The data support on backstage, including the generation of menu module, the recording of menu module, the button of menu module Deng, modularity control ivr menu.

The problem of speech form is described and changes into textual form by step S202, the text techniques that turned by voice；

As it is shown on figure 3, this process can pass through ASR (Automatic Speech Recognition, automatic language Sound identification technology), or other speech recognition technology, the problem of speech form is described and changes into text The problem of form describes.

Step S203, the problem of textual form is described carry out error correction；

Turn, by voice, the problem of textual form that text techniques obtains and describe relatively rough, such as: can be by " remaining sum is precious " changes into " remaining sum guarantor ", need to carry out error correction；

Can set up the basic database for text error correction during this, record everyday words, keyword and Common mistake, it is simple to quickly error correction；

Realize as it is shown on figure 3, this process can pass through AGAP (algorithm platform).

Step S204, problem to user profile precisely identify；

Identify that the problem of the textual form after error correction describes, it is achieved technical ability group belonging to problem types and problem Location；

In the process, can first determine the keyword in problem description, more pre-by described keyword search The mode of the data base first arranged realizes；

As it is shown on figure 3, the problem inputting user describes the process positioned can pass through Csrobot (machine Device people supports platform) realize；

Failing to orient problem types, then directly this problem can be distributed to all-round technical ability group, or Person sets up transfer technical ability group, transfer technical ability group manually receive calls, if transfer technical ability group can solve user Problem is the most directly settled a dispute by the parties concerned themselves customer problem, if can not solve, it is determined that go out customer problem type and transfer To the most special technical ability group；

The hot line outsourcing intelligence shunting model based on adaptive learning that step S205, the use present invention propose, Demand to user carries out shunting prediction；

The shunting model effect of described intelligence is based on real time data and historical data, user's demand is distributed to In Duo Jia outsourcing company one, makes user's demand quickly be processed, and supports business rules.

Have chosen following several eigenvalue from real time data and historical data, X is the final of our decision-making Foundation, further feature is that prediction X provides and supports, as it is shown on figure 3, eigenvalue is by can be by setting up one Csmonitor (service monitoring support platform) monitors acquisition, and this platform can also be shunting decision-making simultaneously The host of model；

X: final decision foundation

S: outsourcing company processes the ability of current demand the soonest.

Incoming call amount in the C:100 second.(time interval can set, by fixed for the time being for this time in the application Being 100 seconds, the time interval 100-150 second is optimal)

R: number of users of currently queuing up.

The sum of N: customer service.

F: the number of idle customer service.

T: the current amount of picking up.

P: day estimates in advance.

α, β, γ, δ are business importance parameter.

Divergence Accordance function is as follows:

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T)$ Formula (1)

Function is explained: 1) C with S is inversely proportional to, and in i.e. 100 seconds, incoming call amount C is the highest, and outsourcing company is not Coming in a period of time, the ability processing current demand the soonest is the most weak.Because the incoming call accessed in 60 seconds The customer service resource that amount takies the most just discharges.For solving the different problem of customer service number of each company, Incoming call amount C is done normalization operate, by incoming call amount divided by outsourcing company customer service number (C/N), finally, The value of C/N is the biggest, and inside and outside following a period of time, the ability of Bao Zheng department process active user's demand is the most weak, Otherwise it is the strongest.

2) R with S is inversely proportional to, and i.e. current queue number is the highest, then outsourcing company is in following a period of time The ability processing current demand the soonest is the most weak, otherwise the strongest

3) F to S is directly proportional, i.e. the number of current idle customer service is the biggest, then outsourcing company is following one The ability processing user's demand in the section time is the strongest, otherwise the most weak.

Therefore, $S=\frac{F}{\left(\frac{C}{N}\right)*R}=\frac{N*F}{C*R}$ Formula (2)

Although formula (2) has met business demand, but in business, C/N, R, F this 3 The business importance of individual factor is different, and the effect of these 3 factors is considered as identical by formula (2). For solving this problem, by formula mutation it is:

$\ln S=ln\frac{N*F}{C*R}=ln\frac{N}{C}+ln\frac{1}{R}+\ln F$ Formula (3)

Addition importance parameter:

$\ln S=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F$ Formula (4)

In formula (4), α, beta, gamma is the business importance journey of these 3 factors of C/N, R, F respectively Degree.

Abnormal conditions consider: when C, R, F are 0, take 1, represent this factor and be not involved in the assessment of S.

Business rules considers (adding two factors): P-T is the biggest, represents that the amount of picking up is estimated more from pre-now Far, need to preferentially be split, X should be the biggest.

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T)$ Formula (5)

Business is explained: α, the effect of these 3 parameter regulation models of beta, gamma, unrelated with business rules, its value Can be gradually stable.The value of δ is defaulted as 0, represents and does not consider business rules, only when the operation result of model Time bigger with business rules gap, just can regulate δ, the result of regulation is to make the currently amount of picking up estimate than day Amount differs bigger outsourcing company and is more likely assigned to user's request, but does not ensure to comply fully with and estimate.

Divergence Accordance: lnS is the biggest, outsourcing company processes that the ability of current demand is the strongest (does not consider business the soonest In the case of industry rule), on the contrary the most weak.

About the determination of importance parameter, in addition to passing through to preset and determining, also two kinds of more excellent determination sides Formula, including supervised learning and the mode of adaptive learning.

First illustrating the mode of supervised learning, supervised learning refers to utilize the sample of one group of known class The parameter of this adjustment grader so that it is reach the process of required properties, also referred to as supervised training or have religion Teacher learns, and supervised learning is the machine learning task that the training data from labelling infers a function.Instruction Practice data and include a set of training example.In supervised learning, each example be by an input object and One desired output valve (also referred to as supervisory signals) composition.Supervised learning algorithm is to analyze this training number According to, and produce a function inferred.

In this application, training data has a F, C, R, N, Label (1), Label (2), Label (3), Label (4), wherein, it is the artificial scoring that i-th processes object that Label (i) represents shunting result.

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T)$

$L={\mathrm{\Σ}}_{i=1}^{n}Score(fi,ci,ri,ni)$ (object function)

Score (fi, ci, ri, ni) is the score of the i-th data shunting result.

Wherein, the target of training is to make target function value L maximum.

Next to that by the way of adaptive learning, adaptive learning typically refers to providing corresponding in study The environment of study, example or field domain, by learner from study in find sum up, ultimately form Theory the mode that can independently solve learning of problem.Adaptive learning in this programme refers to: can be according to mesh The quality of mark result, regulates self model parameter, makes the result of decision more excellent.

In this application, the initial value of α, β, γ can be set to 1, step-length c is set to 0.01, with The value of α, β, γ carries out by machine+,-operation, training natural law is set to U, the user of statistics every day is flat In equal waiting time and process object, the average treatment amount of each processing unit is as evaluation criteria, if User's average waiting time length ratio was short for upper one day, and the average treatment amount of each processing unit is more than upper one day, Then parameter is evolved successfully, when training natural law is equal to U, stops study, and U can be arranged as the case may be.

Step S206, predicting the outcome according to intelligence shunting model, be diverted to user's demand outside correspondence Bao Zheng department；

User's demand is branched to the middle outsourcing company determined that predicts the outcome, completes branching process and complete, or Predicting the outcome described in person is only a reference, also can comprehensively other factors, the outsourcing such as having in reality Company may priority not be the highest, but sequence is the most earlier, and the said firm belongs to together with this user simultaneously One area, it is contemplated that the factors such as dialect, it is also possible to this user's demand is distributed to the said firm's process.

The embodiment of the present application three also proposed a kind of shunting device, as shown in Figure 4, and including:

Receiver module 41, is used for receiving user's request；

Judge module 42, the content for asking based on described user judges the type of customer problem；

Prediction module 43, for type based on described customer problem and shunting model, please to described user Ask and carry out shunting prediction；

Diverter module 44, for result based on described shunting prediction, branches to phase by described user request The process object answered.

Described judge module, specifically includes:

Receive described user request, it is thus achieved that the customer problem of speech form describes；

The customer problem of described speech form is described the customer problem description changing into textual form；

Customer problem description to described textual form is modified；

Identify that the customer problem of revised textual form describes, determine the type of customer problem.

Described prediction module, specifically includes:

Type based on described customer problem, determines alternative process object；

Described alternative process object is carried out prioritization；

Determine that described shunting predicts the outcome based on described prioritization result；

Described prediction module carries out prioritization to described alternative process object, specifically includes:

Based on the priority parameters calculated, described alternative process object is carried out prioritization；

Wherein, described priority parameters is

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T);$

Wherein, processing user in S is the described process object one section of Preset Time after current time please The quantized value of the ability asked；C be current time before one section of Preset Time in described process object receive use The quantity of family request；R is the quantity of the pending user request of described process object；N is described process The quantity of processing unit in object；F is the quantity of idle processing unit in described process object；T is described Process object and the most receive the quantity of user's request；P is that described process object receives estimating of user's request Amount；α, β, γ, δ are parameter.

The determination mode of described parameter, particularly as follows:

Determine by presetting；And/or,

Determined by the process of supervised learning；And/or,

Determined by the process of adaptive learning.

Compared with prior art, the application achieves quickly and shunts accurately, improves and processes object Treatment effeciency, branching process is intelligent, save a large amount of human resources, it is to avoid the error of manual operation, Reduce failure risk on line, and the present invention can process user not only according to process object and ask to shunt, User can be handled well always according to outsourcing company to ask to shunt, improve Consumer's Experience, enhance use The family sense of trust to company.

Through the above description of the embodiments, those skilled in the art is it can be understood that arrive this Shen Please be realized by hardware, it is also possible to the mode adding necessary general hardware platform by software realizes. Based on such understanding, the technical scheme of the application can embody with the form of software product, and this is soft Part product can be stored in a non-volatile memory medium (can be CD-ROM, USB flash disk, portable hard drive Deng) in, including some instructions with so that a computer equipment (can be personal computer, service Device, or the network equipment etc.) each implements the method described in scene to perform the application.

It will be appreciated by those skilled in the art that accompanying drawing is a schematic diagram being preferable to carry out scene, in accompanying drawing Module or flow process not necessarily implement necessary to the application.

It will be appreciated by those skilled in the art that the module in the device implemented in scene can be according to implementing scene Describe and carry out being distributed in the device implementing scene, it is also possible to carry out respective change and be disposed other than this enforcement In one or more devices of scene.The module of above-mentioned enforcement scene can merge into a module, it is possible to To be further split into multiple submodule.

Above-mentioned the application sequence number, just to describing, does not represent the quality implementing scene.

The several scenes that are embodied as being only the application disclosed above, but, the application is not limited to This, the changes that any person skilled in the art can think of all should fall into the protection domain of the application.

Claims (10)

1. a shunt method, it is characterised in that the method comprises the following steps:

Receive user's request；

The content asked based on described user judges the type of customer problem；

Type based on described customer problem and shunting model, carry out shunting prediction to described user request；

Result based on described shunting prediction, branches to process accordingly object by described user request.

2. the method for claim 1, it is characterised in that described ask based on described user in Hold the type judging customer problem, specifically include:

Receive described user request, it is thus achieved that the customer problem of speech form describes；

The customer problem of described speech form is described the customer problem description changing into textual form；

Customer problem description to described textual form is modified；

Identify that the customer problem of revised textual form describes, determine the type of customer problem.

3. the method for claim 1, it is characterised in that described class based on described customer problem Type and shunting model, carry out shunting prediction to described user request, specifically include:

Type based on described customer problem, determines alternative process object；

Described alternative process object is carried out prioritization；

Determine that described shunting predicts the outcome based on described prioritization result；

4. method as claimed in claim 3, it is characterised in that described described alternative process object is entered Row major level sorts, and specifically includes:

Based on the priority parameters calculated, described alternative process object is carried out prioritization；

Wherein, described priority parameters is

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T);$

Wherein, processing user in S is the described process object one section of Preset Time after current time please The quantized value of the ability asked；C be current time before one section of Preset Time in described process object receive use The quantity of family request；R is the quantity of the pending user request of described process object；N is described process The quantity of processing unit in object；F is the quantity of idle processing unit in described process object；T is described Process object and the most receive the quantity of user's request；P is that described process object receives estimating of user's request Amount；α, β, γ, δ are parameter.

5. method as claimed in claim 4, it is characterised in that the determination mode of described parameter, specifically For:

Determine by presetting；And/or,

Determined by the process of supervised learning；And/or,

Determined by the process of adaptive learning.

6. a shunting device, it is characterised in that including:

Receiver module, is used for receiving user's request；

Judge module, the content for asking based on described user judges the type of customer problem；

Prediction module, for type based on described customer problem and shunting model, asks described user Carry out shunting prediction；

Diverter module, for result based on described shunting prediction, branches to corresponding by described user request Process object.

7. equipment as claimed in claim 1, it is characterised in that described judge module, specifically includes:

Receive described user request, it is thus achieved that the customer problem of speech form describes；

The customer problem of described speech form is described the customer problem description changing into textual form；

Customer problem description to described textual form is modified；

Identify that the customer problem of revised textual form describes, determine the type of customer problem.

8. equipment as claimed in claim 1, it is characterised in that described prediction module, specifically includes:

Type based on described customer problem, determines alternative process object；

Described alternative process object is carried out prioritization；

Determine that described shunting predicts the outcome based on described prioritization result；

9. equipment as claimed in claim 3, it is characterised in that described prediction module is to described alternative place Reason object carries out prioritization, specifically includes:

Based on the priority parameters calculated, described alternative process object is carried out prioritization；

Wherein, described priority parameters is

$X=\ln S+\mathrm{\δ}ln(P-T)=\mathrm{\α}ln\frac{N}{C}+\mathrm{\β}ln\frac{1}{R}+\mathrm{\γ}\ln F+\mathrm{\δ}ln(P-T),(P>T);$

Wherein, processing user in S is the described process object one section of Preset Time after current time please The quantized value of the ability asked；C be current time before one section of Preset Time in described process object receive use The quantity of family request；R is the quantity of the pending user request of described process object；N is described process The quantity of processing unit in object；F is the quantity of idle processing unit in described process object；T is described Process object and the most receive the quantity of user's request；P is that described process object receives estimating of user's request Amount；α, β, γ, δ are parameter.

10. equipment as claimed in claim 4, it is characterised in that the determination mode of described parameter, tool Body is:

Determine by presetting；And/or,

Determined by the process of supervised learning；And/or,

Determined by the process of adaptive learning.