CN110399995A

CN110399995A - Waybill complaint handling method, apparatus, equipment and its storage medium

Info

Publication number: CN110399995A
Application number: CN201810364654.0A
Authority: CN
Inventors: 刘琼; 陈佳琦; 席怡雯; 王本玉; 金晶
Original assignee: SF Technology Co Ltd
Current assignee: SF Technology Co Ltd; SF Tech Co Ltd
Priority date: 2018-04-20
Filing date: 2018-04-20
Publication date: 2019-11-01

Abstract

This application discloses waybill complaint handling method, apparatus, equipment and its storage mediums.This method comprises: obtain the related data of waybill to be predicted, the related data of waybill to be predicted belongs to non-signs for type in person；Data processing is carried out to the related data of waybill to be predicted；Result after data processing is input to the waybill constructed in advance and complains model, exports the complaint probability value of waybill to be predicted, waybill, which complains model, to be obtained using machine learning algorithm training study；And based on complaint probability value service strategy corresponding with the comparison result of preset threshold value adjustment waybill to be predicted.According to the technical solution of the embodiment of the present application, by predicting the complaint probability value of waybill, and it is directed to the high waybill for complaining probability and personalized service strategy is provided, to significantly reduce customer complaint probability, the user experience is improved spends.

Description

Waybill complaint handling method, apparatus, equipment and its storage medium

Technical field

Present application relates generally to technical field of information processing, and in particular to the technology of data mining technology processing logistics information Field more particularly to waybill complaint handling method, apparatus, equipment and its storage medium.

Background technique

With the development of logistic industry, express waybill amount is increased rapidly, adjoint and raw subscriber data data, is had very big Utility value.In the scene that waybill sends part, to avoid sending part article from being detained, improving and send part efficiency, it often will appear non- The case where people signs for, for example, household withholds, gatekeeper withholds, rich nest cabinet, e stack etc..It is non-sign in person event occur after, It is possible that the Reference Group such as the relatives of addressee or addressee, room-mate trigger complaint event.

Currently, this kind of complaint event can not be predicted in advance, and countermeasure with no personalization, cause user experience compared with Difference, satisfaction are lower.

Summary of the invention

In view of drawbacks described above in the prior art or deficiency, it is intended to provide that a kind of sign in person can be more under scene non- Accurately prediction waybill is complained the technical solution of probability.

In a first aspect, the embodiment of the present application provides a kind of waybill complaint handling method, this method comprises:

Obtain the related data of waybill to be predicted, the related data of waybill to be predicted belongs to non-signs for type in person；

Data processing is carried out to the related data of waybill to be predicted；

Result after data processing is input to the waybill constructed in advance and complains model, the complaint for exporting waybill to be predicted is general Rate value, wherein waybill, which complains model, is obtained using machine learning algorithm training study；And

Based on complaint probability value service strategy corresponding with the comparison result of preset threshold value adjustment waybill to be predicted.

Second aspect, the embodiment of the present application provide a kind of waybill complaint handling device, which includes:

First acquisition unit, for obtaining the related data of waybill to be predicted, the related data of waybill to be predicted belongs to non- I signs for type；

First data processing unit carries out data processing for the related data to waybill to be predicted；

Probability prediction unit complains model, output for the result after data processing to be input to the waybill constructed in advance The complaint probability value of waybill to be predicted, waybill, which complains model, to be obtained using machine learning algorithm training study；And

Developing Tactics unit, for based on the comparison result adjustment waybill pair to be predicted for complaining probability value and preset threshold value The service strategy answered.

The third aspect, the embodiment of the present application provide a kind of computer equipment, including memory, processor and are stored in On memory and the computer program that can run on a processor, the processor realize such as the embodiment of the present application when executing the program The method of description.

Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer journey Sequence, the computer program are used for:

The method as described in the embodiment of the present application is realized when the computer program is executed by processor.

Waybill complaint handling method provided by the embodiments of the present application, by predict waybill complaint probability value, and by its with Preset threshold is compared, and provides targetedly service strategy, to solve the problems, such as prior art complaint handling lag, is had Customer complaint probability is reduced to effect, the user experience is improved spends.The embodiment of the present application is also further by establishing user's community Improve the precision of prediction of prediction model.

Detailed description of the invention

By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other Feature, objects and advantages will become more apparent upon:

Fig. 1 shows the flow diagram of the waybill complaint handling method of the embodiment of the present application proposition；

Fig. 2 shows the flow diagrams that the building waybill that the another embodiment of the application provides complains model method；

Fig. 3 shows the flow diagram of another waybill complaint handling method provided by the embodiments of the present application；

Fig. 4 shows the exemplary block diagram of waybill complaint handling device 400 provided by the embodiments of the present application；

Fig. 5 shows the exemplary block diagram that building waybill provided by the embodiments of the present application complains model equipment 500；

Fig. 6 shows the exemplary block diagram of another waybill complaint handling device 600 provided by the embodiments of the present application；

Fig. 7 shows the structural schematic diagram for being suitable for the computer system for being used to realize the embodiment of the present application.

Specific embodiment

The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to Convenient for description, part relevant to invention is illustrated only in attached drawing.

It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

There are a variety of possible scenes complained in logistic industry, such as sign for caused complaint field in person there may be non- Scape.For the scene, the embodiment of the present application, which proposes one kind, can predict the non-skill for signing for the complained probability value of waybill in person Art scheme.

Referring to FIG. 1, Fig. 1 shows the flow diagram of the waybill complaint handling method of the embodiment of the present application proposition.

As shown in Figure 1, this method comprises:

Step 110, the related data of waybill to be predicted is obtained, wherein the related data of waybill to be predicted belongs to non-label in person Receive type.

In the embodiment of the present application, when waybill monitoring platform obtains new waybill event, signed for by obtaining new waybill identification Mark determines whether the new waybill belongs to the non-waybill signed in person, is used as if the waybill that right and wrong are signed for be predicted Object.

Step 120, data processing is carried out to the related data of waybill to be predicted.

In the embodiment of the present application, certain parameters in waybill data are subjected to special data processing, can reduce model Site code information in the difficulty of building, such as waybill data can be mapped as corresponding by average coding mode Encoded radio.Site code information is constructed into an average Code Mapping Tables in advance by a large amount of historical data, by this Mapping table establishes site code information and according to the association between the corresponding encoded radio of average coding mode generation.

After obtaining the related data of new waybill, the site code information in related data by obtaining the new waybill, Then the average Code Mapping Tables constructed in advance are searched, available encoded radio corresponding with the site code information is real Now to the data processing of the related data of new waybill.

Step 130, the result after data processing is input to the waybill constructed in advance and complains model, exports waybill to be predicted Complaint probability value, wherein waybill complain model be using machine learning algorithm training study obtain.

In the embodiment of the present application, by the way that by step 120, treated that waybill that waybill data are input to be constructed in advance is complained In model, the complained probability value of the waybill is exported.Wherein, it is using machine learning algorithm that the waybill constructed in advance, which complains model, Training study obtains, which complains model to include at least two submodels, and each submodel is all made of machine learning algorithm Training study obtains, such as the first submodel, using the training study of the first machine learning algorithm, the second submodel uses the second machine The training study of device learning algorithm, two submodels for then obtaining training study, proportionally combination producing waybill complains mould Type.Wherein, the first machine learning algorithm and the second machine learning algorithm, such as can be logistic regression algorithm, random forest calculation Method and other machines learning algorithm.Preferably, the first machine learning algorithm is logistic regression algorithm, and the second machine learning is calculated Method is random forests algorithm.

It is complained in model process in building waybill, by extracting the characteristic parameter for influencing the complained various dimensions of waybill come table Model is levied, thus the mapping relations between whether construction feature parameter and waybill are complained.Wherein influence the complained spy of waybill Parameter is levied, such as can be waybill dimension, user's dimension, send the features such as part dimension.Wherein, user's dimension is as influence waybill quilt One dependent variable of the result of complaint, it is particularly important on the complained influence of waybill.Cause the field complained in non-waybill of signing in person Under scape, there are incidence relations between the non-related personnel's data signed in person and waybill complaint.For example, with same standardized address It is limited, multiple personnel of same standardized address are relatively common on behalf of signing for.The embodiment of the present application, it is contemplated that this great influence The factor, definition can be built into user's community with multiple users that standardized address is reference, pass through machine learning algorithm Excavate user's community information and influence waybill it is complained between incidence relation.

The embodiment of the present application can will predict the preposition processing of calling information by introducing user's community information, and further Promote the prediction accuracy of prediction model.

Step 140, based on the service corresponding with the comparison result of preset threshold value adjustment waybill to be predicted of complaint probability value Strategy.

After the embodiment of the present application obtains the predicted value that waybill complains model output, using the predicted value and preset threshold value into Row compares, and if it is greater than the threshold value, then shows that waybill to be predicted belongs to high probability and complains waybill, complains waybill to mention for high probability For corresponding service strategy.

Wherein, the value of threshold value can for example be determined by percentile method, can also be according to practical application scene Setting, the value mode of the threshold value is the ratio determined based on actual intervention capacity.

The embodiment of the present application complains model that can obtain the complained probability of waybill in advance by the waybill constructed in advance Value to solve the problems, such as to complain event information lag in the prior art, and takes clothes corresponding with high probability complaint waybill The tactful Experience Degree to promote user of business, improves service quality.

Further, step 130 predicts fortune to be predicted by the waybill complaint model constructed in advance in the embodiment of the present application Singly complained probability value, waybill complain model to be referred to the method for Fig. 2 description to pre-establish.

Referring to FIG. 2, Fig. 2 shows the building waybills that the another embodiment of the application provides, and the process of model method to be complained to show It is intended to.

As shown in Fig. 2, this method comprises:

Step 210, the first machine learning model is obtained using the training study of the first machine learning algorithm；

Step 220, the second machine learning model is obtained using the training study of the second machine learning algorithm；

Step 230, it combines the first machine learning model and the second machine learning model to obtain waybill complaint according to weight Model.

In the embodiment of the present application, using machine learning algorithm training waybill data, obtains waybill and complain model, for predicting The complained probability of waybill.Wherein the process of training pattern includes three parts.First part, using the first machine learning algorithm Training study obtains the first machine learning model；Second part obtains the second machine using the training study of the second machine learning algorithm Device learning model；Part III is then proportionally to combine first part and second part result.

Wherein, before step 210, this method can also include:

Step 210a obtains history waybill data, which belongs to non-signs for type in person；

Step 210b carries out data processing to history waybill data；

Result after data processing is divided into training dataset and test data set according to preset ratio by step 210c.

First machine learning model is being obtained using the training study of the first machine learning algorithm and is using the second machine learning Before algorithm training study obtains the second machine learning model, needs to obtain and largely belong to the non-history waybill number signed in person According to, then data processing is carried out to history waybill data.Then training dataset and test data set are established using method of random sampling, Final waybill prediction model is obtained based on training dataset and test data set training study.

History waybill data are pre-processed in step 210b, such as can be by carrying out data to the wide table of basic data The operations such as cleaning are realized.Data cleansing operation for example may include duplicate removal processing, to null value, the data without practical significance carry out The operation such as data conversion, duplicate removal processing can reject repeated data；To null value, the data without practical significance carry out data conversion can To be realized by filling up missing values.Different pretreatment modes is used with specific reference to different data types.The application is to pre- The mode of processing is not especially limited.After completing cleaning treatment, training dataset and verifying number are established using method of random sampling According to collection, such as 7: 3.

Corresponding machine learning model is obtained using machine learning algorithm training using training dataset, the machine learning mould Type is for predicting the whether potential complained risk of the non-waybill signed in person.Machine learning model provided by the embodiments of the present application Including at least two parts, first part obtains the first machine learning model using the training study of the first machine learning algorithm, the Two parts obtain the second machine learning model using the training study of the second machine learning algorithm.

Wherein, step 210 obtains the first machine learning model using the training study of the first machine learning algorithm, such as may be used also To include:

Step 2101, the first machine learning mould is obtained according to the training study of the first machine learning algorithm using training dataset Type, the first machine learning model are that the characteristic parameter that the first machine learning algorithm extracts and non-waybill of signing in person are complained probability Mapping relations between value, this feature parameter include at least waybill dimension, user's dimension, send part dimension, wherein user's dimension packet User's community information is included, sending part dimension includes the corresponding encoded radio of site code information.

Step 2102, test data set is inputted into the first machine learning model and exports the first probability value.

The embodiment of the present application, can be from the complained main feature of influence waybill data by the first machine learning algorithm Main feature is extracted, to construct the first machine learning model, wherein the main feature extracted includes at least waybill dimension, Yong Huwei Spend, send part dimension etc..

Wherein, waybill dimension include the pulling between the time receiving of waybill, waybill weight, the area addressee great Qu, Ji Jian great, situation of supporting value, Volume (or waybill volume), waybill expense, payment type, the support that frangible situation, support post object post species type, waybill type etc..

Wherein, user's dimension includes user's community information.User's community information refer to user retain addressee information/ Standardized address in sender's information is reference, constructs the incidence relation of multiple users relevant to the standardized address.Example Such as, the standardization posting address that individual traveler user A, B, C are retained is certain cell building unit room (such as number), then fixed Adopted individual traveler user A, B, C are a community.Alternatively, the corresponding business address of corporate user, multiple users use identical standard Change posting address, then multiple users is defined as a community.

Wherein, sending part dimension includes the corresponding encoded radio of site code information.The embodiment of the present application is obtained by historical data Site code information is taken, site code information is then used into average encryption algorithm, the flat of site code information is calculated Equal encoding value.It calculates separately multiple and different site code informations to obtain its corresponding average encoding value to construct averagely Number encoder mapping table.

The first machine learning model is constructed according to the main feature of extraction and its parameter, which is the first machine learning The characteristic parameter of extraction and the non-mapping relations signing for waybill in person and being complained between probability value.

Then, it is input to that non-waybill of signing in person is obtained in the first machine learning model is complained using test data set First probability value, and the first AUC value is calculated based on the first probability value.

Further, step 220, the second machine learning model is obtained using the training study of the second machine learning algorithm, such as Can also include:

Step 2201, the second machine learning mould is obtained according to the training study of the second machine learning algorithm using training dataset Type, wherein the second machine learning model is waybill dimension, user's dimension, part dimension and non-waybill of signing in person is sent to be complained generally Mapping relations between rate value, user's dimension include user's community information, and sending part dimension includes that site code information is corresponding Encoded radio；

Step 2202, test data set is inputted into the second machine learning model and exports the second probability value.

The embodiment of the present application signs for whether waybill is complained to become in response by the second machine learning algorithm, with non-in person Amount, waybill dimension, user's dimension send part dimension to carry out data training for dependent variable, obtain waybill dimension, user's dimension, send part Dimension and it is non-sign in person waybill it is whether complained between mapping relations, wherein waybill dimension, user's dimension, send part dimension and The characteristic parameter of first part is identical.

Further, step 230, it combines the first machine learning model and the second machine learning model and is transported according to weight Single-throw tells model, can also include:

First probability value is summed to obtain with the second probability value multiplied by weight coefficient multiplied by the difference of numerical value 1 and weight coefficient Waybill complains model, wherein weight coefficient is determined according to the AUC maximization principle of test data set.

Test data set is inputted the first machine learning model and predicted by the embodiment of the present application, available test number According to the first prediction probability value for concentrating each history waybill complained, the probability value is related to the first machine learning model.It will test Data set inputs the second machine learning model and is predicted, available test data concentrates complained second of each history waybill Prediction probability value, the probability value are related to the second machine learning model.

After obtaining the first probability value and the second probability value, using the first probability value multiplied by weight coefficient and the second probability Value is summed multiplied by the difference of numerical value 1 and weight coefficient, obtains the complained probability value of waybill.

Wherein, weight coefficient is determined according to the AUC maximum principle of test data set.Weight is determined according to AUC maximum principle Coefficient, can be by assigning different values (wherein, the value range of w is the numerical value between 0 to 1) to weight coefficient w, in terms of Calculating test data concentrates each history waybill to complain the corresponding AUC of result of model output to determine weight coefficient by waybill.

For example, assigning 0.3 value to w for the first time, then the second probability value is added multiplied by numerical value multiplied by the first probability value by w The difference of 1 and w obtains each history waybill by waybill complaint model output as a result, and calculating the corresponding AUC of the result.

Then, by adjusting the value of w, for example, w assignment 0.6, calculates each history waybill of each history waybill by fortune again Single-throw tell the output of total model as a result, and calculating the corresponding AUC of the result.

The step of above-mentioned assignment calculates is repeated, finally more all AUC, determine the value conduct of the corresponding w of maximum AUC Weight coefficient.

On that basi of the above embodiments, the application also proposed a kind of processing method that waybill is complained for high probability.Please The flow diagram of another waybill complaint handling method provided by the embodiments of the present application is shown with reference to Fig. 3, Fig. 3.

As shown in figure 3, the method comprising the steps of 310-330, identical as method and step 110-130, specifically describe referring to step The description of rapid 110-130.

Step 340, based on the service corresponding with the comparison result of preset threshold value adjustment waybill to be predicted of complaint probability value Strategy.The step can also include:

Step 3401, it is compared using complaint probability value with preset threshold value；

Step 3402, if it is greater than the threshold value, then based on whether in the presence of with the related data of waybill to be predicted is associated pushes away Recording mark is sent to push corresponding service strategy.

The embodiment of the present application judges whether corresponding waybill belongs to the throwing of high probability by probability value compared with threshold value Waybill is told, if it is, further judging the addressee information for including in waybill/sender's information with the presence or absence of push record mark Note, is then based on this judging result to push corresponding service strategy, to promote the Experience Degree of user.

Wherein, based on whether in the presence of right therewith to push with the associated push recording mark of the related data of waybill to be predicted The service strategy answered can also include:

Step 3402a obtains the first identifier information in the related data of waybill to be predicted；

Step 3402b marks list based on first identifier information searching history relevant to the first identifier information；

Step 3402c judges in history label list with the presence or absence of push recording mark；

Step 3402d pushes corresponding service strategy based on the judging result of push recording mark.

The embodiment of the present application, after judging that waybill to be predicted belongs to the complaint waybill of high probability, further to be predicted First identifier information is obtained in the related data of waybill, which for example can be addressee/sender connection Mode or monthly closing entry account etc..For example, the first identifier information of individual traveler user can be the phone number of addressee or post part The phone number of people；The first identifier information of corporate user for example can be monthly closing entry account etc..

It whether there is by position corresponding with first identifier information in the label list of first identifier information searching history and push away Send recording mark.Wherein, the list of history label can in tabular form or database mode.The push recording mark is for example It can be and pushed, or do not push.Perhaps using the push of short message, perhaps wechat pushes push record or other are When communication mode push etc..

If finding push recording mark corresponding with first identifier information in history label list, judge to pre- It surveys addressee information/sender's information in the related data of waybill and belongs to non-service object for the first time.

If history label list in do not find push recording mark corresponding with first identifier information, judge to Predict that addressee information/sender's information belongs to service object for the first time in the related data of waybill.

Further directed to service object for the first time and non-service object for the first time, user can be improved by different service strategies Experience Degree.

For service object for the first time, corresponding service strategy is pushed based on the judging result of push recording mark Step can also include:

If there is no push recording mark, then first service strategy is pushed.

The embodiment of the present application, by monitoring whether waybill to be predicted by correct-distribute label executes Push Service strategy to trigger Movement.If monitoring that corresponding article or quick despatch have completed correct-distribute, corresponding functional module is triggered, such as push Module pushes first service strategy to user.Wherein, first service strategy includes the combination of push mode and push content.It pushes away Send mode for example can be using wechat, short message, other instant communication modes etc., push content, which for example can be to user, pushes certain Waybill correct-distribute, and by the prompt informations content such as someone's allograph.It is transported in time by first service strategy to user feedback Single signs for result.

And after pushing first service strategy to user, this method can also include:

Step 350, feedback letter corresponding with addressee information in the related data of waybill to be measured/sender's information is monitored Breath, and second identifier information is generated according to feedback information.

After pushing first service strategy, continues monitoring and believe with addressee corresponding in the related data of waybill to be predicted Breath/sender's information feedback information, and second identifier information is generated according to the feedback information.Wherein, second identifier information is used In indicating whether and the corresponding feedback information of push recording mark.The second identifier information is also used as selection service plan The slightly reference of push-mechanism.

For example, system does not receive the feedback of addressee after addressee receives the first service policy message of push Information then indicates second identifier information, such as short message, wechat text information using the push mode of first service strategy.When After addressee receives the first service policy message of push, system receives the complaint event of addressee, then using non-first clothes Tactful push mode be engaged in indicate the text informations such as second identifier information, such as non-short message, non-wechat.

For non-service object for the first time, corresponding service plan is pushed based on the judging result of push recording mark Slightly, can also include:

If there is push recording mark, obtains addressee information/sender in the related data with waybill to be predicted and believe The corresponding second identifier information of manner of breathing, the second identifier information are used to indicate whether there is feedback corresponding with push recording mark Information；

First service strategy or second service strategy are pushed based on second identifier information, the second service strategy and described the One service strategy is different.

The embodiment of the present application, addressee information/sender's information belongs to non-in the related data for judging waybill to be predicted For the first time after service object, the second identifier information in the related data of waybill to be predicted is further obtained, then, according to acquisition Second identifier information select corresponding service strategy.The second identifier information is generated according to the feedback information of user, as choosing Select the reference frame of service strategy.I.e. if second identifier information is used to indicate, there is no feedbacks corresponding with push recording mark Information then selects first service strategy.If second identifier information is used to indicate in the presence of feedback corresponding with push recording mark Information then selects second service strategy.

It is found that second identifier information for example can be short message/non-short message, wechat/non-wechat etc. in the description of step 350 Form.

When second identifier information is short message/wechat, then it represents that addressee/sender received service in waybill to be predicted The PUSH message of strategy, and complaint event is not present in the service strategy pushed for before, it can be to for waybill to be predicted User pushes identical service strategy, to reduce the complaint probability of user.

When second identifier information is non-short message/non-wechat, then it represents that addressee/sender received in waybill to be predicted The PUSH message of service strategy, and there are complaint events for the service strategy pushed for before, then need for waybill to be predicted Adjust service strategy, using from service strategies different before, come reduce user complaint probability.

Wherein, first service strategy includes the combination of push mode and push content.Push mode can for example use micro- Letter, short message, other instant communication modes etc., push content, which for example can be to user, pushes the correct-distribute of certain waybill, and by certain The prompt informations content such as someone allograph.Result is signed for user feedback waybill by first service strategy.

Wherein, second service strategy includes the combination of push mode, push content, other push objects.Push mode example Wechat, short message, other instant communication modes can be such as used, push content, which for example can be to user, has pushed certain waybill Correct-distribute, and by the contents such as the prompt informations such as someone's allograph content and other service items.Obtaining second identifier information When, also to other Object Push reminder messages, for example, sending part personnel to push reminder message to waybill to be predicted, which disappears Breath need to be signed for for prompting certain waybill this time to deliver by user.Pass through second service strategy signing for user feedback waybill It as a result and other service items, can also be by second service strategy to sending part personnel to submit reminder message.

For example, in the related data for judging waybill to be predicted addressee information/sender's information belong to it is non-for the first time After service object, second identifier information corresponding with addressee information/sender's information is further obtained, if obtaining second Identification information is short message, after receiving certain waybill by correct-distribute information, then pushes the correct-distribute of certain waybill to user, and by so-and-so The prompt informations content such as people's allograph.

If acquisition second identifier information is that non-short message sends part people to this when certain waybill is assigned to and sends part personnel Member's push reminder message, the reminder message content include that certain waybill such as need to sign at the prompt informations in person.And receiving certain waybill quilt After correct-distribute information, the correct-distribute of certain waybill is pushed to user, and believed by the prompts such as someone's allograph and other service items Cease content.Alternatively, if obtaining second identifier information is non-short message, only after receiving certain waybill by correct-distribute information, to user The correct-distribute of certain waybill is pushed, and by the prompt informations content such as someone's allograph and other service items.

The embodiment of the present application improves the service quality that express delivery is delivered, to promote user by personalized service strategy Experience Degree, reduce the non-complaint probability for signing for waybill in person.

It should be noted that although describing the operation of the method for the present invention in the accompanying drawings with particular order, this is not required that Or hint must execute these operations in this particular order, or have to carry out operation shown in whole and be just able to achieve the phase The result of prestige.On the contrary, the step of describing in flow chart can change and execute sequence.Additionally or alternatively, it is convenient to omit certain Multiple steps are merged into a step and executed, and/or a step is decomposed into execution of multiple steps by step.

Based on the above embodiment, the embodiment of the present application also provides a kind of waybill complaint handling devices, referring to FIG. 4, Fig. 4 Show the exemplary block diagram of waybill complaint handling device 400 provided by the embodiments of the present application.The device 400 includes:

First acquisition unit 410, for obtaining the related data of waybill to be predicted, wherein the related data of waybill to be predicted Belong to and non-signs for type in person.

First data processing unit 420 carries out data processing for the related data to waybill to be predicted.

Probability prediction unit 430 complains model for the result after data processing to be input to the waybill constructed in advance, defeated The complaint probability value of waybill to be predicted out, wherein waybill, which complains model, to be obtained using machine learning algorithm training study.

In the embodiment of the present application, by the way that by the first data processing unit 420, treated that waybill data are input to preparatory structure The waybill built is complained in model, and the complained probability value of the waybill is exported.Wherein, it is to use that the waybill constructed in advance, which complains model, Machine learning algorithm training study obtains, which complains model to include at least two submodels, and each submodel is all made of Machine learning algorithm training study obtains, such as the first submodel is using the training study of the first machine learning algorithm, the second submodule For type using the training study of the second machine learning algorithm, two submodels for then obtaining training study proportionally organize symphysis Model is complained at waybill.Wherein, the first machine learning algorithm and the second machine learning algorithm, such as can be logistic regression calculation Method, random forests algorithm and other machines learning algorithm.Preferably, the first machine learning algorithm is logistic regression algorithm, the Two machine learning algorithms are random forests algorithm.

Developing Tactics unit 440, for based on the comparison result adjustment fortune to be predicted for complaining probability value and preset threshold value Single corresponding service strategy.

The embodiment of the present application complains model that can obtain complained general of waybill data in advance by the waybill that constructs in advance Rate to solve the problems, such as to complain event information lag in the prior art, and takes clothes corresponding with high probability complaint waybill The tactful Experience Degree to promote user of business, improves service quality.

Further, the waybill constructed in advance in the embodiment of the present application complains model, referring to FIG. 5, Fig. 5 shows the application The building waybill that embodiment provides complains the exemplary block diagram of model equipment 500.

As shown in figure 5, the device 500 includes:

First model subelement 510, for obtaining the first machine learning mould using the training study of the first machine learning algorithm Type；

Second model subelement 520, for obtaining the second machine learning mould using the training study of the second machine learning algorithm Type；

Subelement 530 is combined, for combining the first machine learning model and the second machine learning model according to weight Model is complained to waybill.

Wherein, before the first model subelement 510, which can also include:

Second obtains subelement 510a, and for obtaining history waybill data, which belongs to non-signs in person Type；

Second data processing subelement 510b, for carrying out data processing to history waybill data；

Ratio cut partition subelement 510c, for the result after data processing to be divided into training dataset according to preset ratio And test data set.

History waybill data are pre-processed in second data processing subelement 510b, such as can be by basic number The operations such as data cleansing are carried out according to wide table to realize.Data cleansing operation for example may include duplicate removal processing, to null value, without reality The data of meaning carry out the operation such as data conversion, and duplicate removal processing can reject repeated data；To null value, without the data of practical significance Carrying out data conversion can be realized by filling up missing values.Different pretreatment sides is used with specific reference to different data types Formula.The application is not especially limited pretreated mode.After completing cleaning treatment, training number is established using method of random sampling According to collection and validation data set, such as 7: 3.

Wherein, the first model subelement 510, such as can also include:

First training submodule 5101, for being obtained using training dataset according to the training study of the first machine learning algorithm First machine learning model, first machine learning model are the characteristic parameter that the first machine learning algorithm extracts and non-label in person The mapping relations that waybill is complained between probability value are received, this feature parameter includes at least waybill dimension, user's dimension, part is sent to be tieed up Degree, user's dimension includes user's community information, and sending part dimension includes the corresponding encoded radio of site code information.

First test submodule 5102 exports the first probability for test data set to be inputted the first machine learning model Value.

Wherein, sending part dimension includes the corresponding encoded radio of site code information.The embodiment of the present application is by from historical data Then site code information is used average encryption algorithm, site code information is calculated by middle acquisition site code information Average encoding value.It calculates separately multiple and different site code informations to obtain its corresponding average encoding value to construct Average Code Mapping Tables.

Further, the second model subelement 520, such as can also include:

Second training submodule 5201, for being learnt using training data set according to the training of the second machine learning algorithm To the second machine learning model, second machine learning model is waybill dimension, user's dimension, sends part dimension and non-label in person The mapping relations that waybill is complained between probability value are received, user's dimension includes user's community information, and sending part dimension includes site generation The corresponding encoded radio of code information；

Second test submodule 5202 exports the second probability for test data set to be inputted the second machine learning model Value.

Further, subelement 530 is combined, is also used to the first probability value multiplied by weight coefficient and the second probability value multiplied by number Value 1 and the difference of weight coefficient sum to obtain waybill complaint model, wherein weight coefficient is maximum according to the AUC of test data set Change principle to determine.

On that basi of the above embodiments, the application also proposed a kind of processing unit that waybill is complained for high probability.Please The exemplary block diagram of another waybill complaint handling device provided by the embodiments of the present application is shown with reference to Fig. 6, Fig. 6.

As shown in fig. 6, the device includes unit 610-630 corresponding with method and step 110-130, specifically describe referring to step The description of rapid 110-130.

Developing Tactics unit 640, for based on the comparison result adjustment fortune to be predicted for complaining probability value and preset threshold value Single corresponding service strategy.Developing Tactics unit 640 can also include:

Comparing subunit 6401, for being compared using complaint probability value with preset threshold value；

Strategy push subelement 6402, is used for if it is greater than the threshold value, then based on whether in the presence of the phase with waybill to be predicted The push recording mark of data correlation is closed to push corresponding service strategy.

Wherein, strategy pushes subelement 6402, can also include:

Third obtains module 6402a, the first identifier information in related data for obtaining waybill to be predicted；

Searching module 6402b, for being based on first identifier information searching history flag column relevant to first identifier information Table；

Second judgment module 6402c, for judging in history label list with the presence or absence of push recording mark；

Tactful pushing module 6402d, for pushing corresponding service based on the judging result of push recording mark Strategy.

The embodiment of the present application, after judging that waybill to be predicted belongs to high probability complaint waybill, further from fortune to be predicted First identifier information is obtained in single related data, which for example can be addressee/sender correspondent party Formula or monthly closing entry account etc..For example, the first identifier information of individual traveler user can be phone number or the sender of addressee Phone number；The first identifier information of corporate user for example can be monthly closing entry account etc..

For service object for the first time, tactful pushing module 6402d can also include:

First push submodule, for pushing first service strategy if there is no push recording mark.

And after pushing first service strategy to user, which can also include:

Feedback processing modules 650, for monitoring and addressee information in the related data of waybill to be measured/sender's information phase Corresponding feedback information, and second identifier information is generated according to feedback information.

After pushing first service strategy, continue monitoring addressee's letter corresponding with the related data of waybill to be predicted Breath/sender's information feedback information, and second identifier information is generated according to the feedback information.Wherein, second identifier information is used In indicating whether and the corresponding feedback information of push recording mark.The second identifier information is also used as selection service plan The slightly reference of push-mechanism.

For non-service object for the first time, tactful pushing module 6402d can also include:

4th acquisition submodule, for obtaining in the related data with waybill to be predicted if there is push recording mark The corresponding second identifier information of addressee information/sender's information, the second identifier information are used to indicate whether to exist and push away Send recording mark corresponding feedback information.

Second push submodule, for being somebody's turn to do based on second identifier information push first service strategy or second service strategy Second service strategy is different from the first service strategy.

It is found that second identifier information for example can be short message/non-short message in the description of feedback processing modules 650, wechat/ The forms such as non-wechat.

It should be appreciated that each in the method that all units or module recorded in device 400-600 are described with reference Fig. 1-3 Step is corresponding.Device 400-600 and wherein included is equally applicable to above with respect to the operation and feature of method description as a result, Unit, details are not described herein.Device 400 can realizes in advance in the browser of electronic equipment or other security applications, can also In the browser or its security application for being loaded into electronic equipment in a manner of through downloading etc..Corresponding units in device 400 can Cooperated with the unit in electronic equipment to realize the scheme of the embodiment of the present application.

Below with reference to Fig. 7, it illustrates the calculating of the terminal device or server that are suitable for being used to realize the embodiment of the present application The structural schematic diagram of machine system 700.

As shown in fig. 7, computer system 700 includes central processing unit (CPU) 701, it can be read-only according to being stored in Program in memory (ROM) 702 or be loaded into the program in random access storage device (RAM) 703 from storage section 708 and Execute various movements appropriate and processing.In RAM703, also it is stored with system 500 and operates required various programs and data. CPU701, ROM702 and RAM 703 is connected with each other by bus 704.Input/output (I/O) interface 705 is also connected to bus 704。

I/O interface 705 is connected to lower component: the importation 706 including keyboard, mouse etc.；It is penetrated including such as cathode The output par, c 707 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.；Storage section 708 including hard disk etc.； And the communications portion 709 of the network interface card including LAN card, modem etc..Communications portion 709 via such as because The network of spy's net executes communication process.Driver 710 is also connected to I/O interface 705 as needed.Detachable media 711, such as Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 710, in order to read from thereon Computer program be mounted into storage section 708 as needed.

Particularly, in accordance with an embodiment of the present disclosure, it is soft to may be implemented as computer for the process above with reference to Fig. 1-3 description Part program.For example, embodiment of the disclosure includes a kind of computer program product comprising be tangibly embodied in machine readable Jie Computer program in matter, the computer program include the program code for executing the method for Fig. 1-3.In such implementation In example, which can be downloaded and installed from network by communications portion 709, and/or from detachable media 711 It is mounted.

Flow chart and block diagram in attached drawing are illustrated according to the system of various embodiments of the invention, method and computer journey The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation A part of one module, program segment or code of table, a part of aforementioned modules, program segment or code include one or more Executable instruction for implementing the specified logical function.It should also be noted that in some implementations as replacements, institute in box The function of mark can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are practical On can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it wants It is noted that the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, Ke Yiyong The dedicated hardware based system of defined functions or operations is executed to realize, or can be referred to specialized hardware and computer The combination of order is realized.

Being described in the embodiment of the present application involved unit or module can be realized by way of software, can also be with It is realized by way of hardware.Described unit or module also can be set in the processor, for example, can be described as: A kind of processor includes first acquisition unit, the first data processing unit and predicting unit.Wherein, these units or module Title does not constitute the restriction to the unit or module itself under certain conditions, for example, first acquisition unit can also be retouched It states as " for obtaining the unit of the non-waybill data to be predicted signed in person ".

As on the other hand, present invention also provides a kind of computer readable storage medium, the computer-readable storage mediums Matter can be computer readable storage medium included in aforementioned device in above-described embodiment；It is also possible to individualism, not The computer readable storage medium being fitted into equipment.Computer-readable recording medium storage has one or more than one journey Sequence, foregoing routine are used to execute the waybill complaint prediction technique for being described in the application by one or more than one processor.

Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic Scheme, while should also cover in the case where not departing from aforementioned invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein Can technical characteristic replaced mutually and the technical solution that is formed.

Claims

1. a kind of waybill complaint handling method, which is characterized in that this method comprises:

Obtain the related data of waybill to be predicted, the related data of the waybill to be predicted belongs to non-signs for type in person；

Data processing is carried out to the related data of the waybill to be predicted；

Result after the data processing is input to the waybill constructed in advance and complains model, exports the throwing of the waybill to be predicted Tell probability value, the waybill, which complains model, to be obtained using machine learning algorithm training study；And

Comparison result based on complaint probability value and the preset threshold value adjusts the corresponding service strategy of the waybill to be predicted.

2. the method according to claim 1, wherein the waybill, which complains model, is instructed using machine learning algorithm Practicing study is included:

First machine learning model is obtained using the training study of the first machine learning algorithm；

Second machine learning model is obtained using the training study of the second machine learning algorithm；

It combines first machine learning model and second machine learning model to obtain waybill complaint model according to weight.

3. according to the method described in claim 2, it is characterized in that, being learnt described using the training of the first machine learning algorithm To before the first machine learning model, the waybill, which complains model, to be obtained using machine learning algorithm training study further include:

History waybill data are obtained, the history waybill data belong to non-signs for type in person；

Data processing is carried out to the history waybill data；

Result after data processing is divided into training dataset and test data set according to preset ratio.

4. according to the method described in claim 3, it is characterized in that, described obtained using the training study of the first machine learning algorithm First machine learning model, comprising:

The first machine learning model, institute are obtained according to first machine learning algorithm training study using the training dataset Stating the first machine learning model is that the characteristic parameter that first machine learning algorithm extracts is complained with non-waybill of signing in person Mapping relations between probability value, the characteristic parameter include at least waybill dimension, user's dimension, send part dimension, wherein described User's dimension includes user's community information, and described to send part dimension include the corresponding encoded radio of site code information；

The test data set is inputted into the first machine learning model and exports the first probability value.

5. according to the method described in claim 4, it is characterized in that, described obtained using the training study of the second machine learning algorithm Second machine learning model, comprising:

The second machine learning model, institute are obtained according to second machine learning algorithm training study using the training dataset State the second machine learning model be waybill dimension, user's dimension, send part dimension and it is non-sign in person the complained probability value of waybill it Between mapping relations, user's dimension includes user's community information, and described to send part dimension include that site code information is corresponding Encoded radio；

The test data set is inputted into the second machine learning model and exports the second probability value.

6. according to the method described in claim 5, it is characterized in that, first machine learning model and second engineering Model is practised to combine to obtain waybill complaint model according to weight, comprising:

By first probability value multiplied by weight coefficient and second probability value multiplied by the difference of numerical value 1 and the weight coefficient Summation obtains the waybill and complains model.

7. method according to claim 1 to 6, which is characterized in that described based on the complaint probability value and pre- If the corresponding service strategy of the comparison result adjustment waybill to be predicted of threshold value include:

It is compared using the complaint probability value with preset threshold value；

If it is greater than the threshold value, then based on whether being marked in the presence of with the associated push record of the related data of the waybill to be predicted Note is to push corresponding service strategy.

8. the method according to the description of claim 7 is characterized in that described based on whether in the presence of the phase with the waybill to be predicted The push recording mark of data correlation is closed to push corresponding service strategy, further includes:

Obtain the first identifier information in the related data of the waybill to be predicted；

List is marked based on first identifier information searching history relevant to the first identifier information；

Judge in the history label list with the presence or absence of push recording mark；

Corresponding service strategy is pushed based on the judging result of the push recording mark.

9. according to the method described in claim 8, it is characterized in that, it is described based on it is described push recording mark judging result come Push corresponding service strategy, comprising:

If there is no the push recording mark, first service strategy is pushed.

10. according to the method described in claim 9, it is characterized in that, this method is also after the push first service strategy Include:

Monitor feedback information corresponding with addressee information/sender's information in the related data of the waybill to be measured, and root Second identifier information is generated according to feedback information.

11. according to the method described in claim 8, it is characterized in that, the judging result based on the push recording mark To push corresponding service strategy, comprising:

If there is the push recording mark, obtains addressee information/sender in the related data with waybill to be predicted and believe The corresponding second identifier information of manner of breathing, the second identifier information are used to indicate whether to exist corresponding with the push recording mark Feedback information；And

First service strategy or second service strategy are pushed based on the second identifier information, the second service strategy and described the One service strategy is different.

12. -11 described in any item methods according to claim 1, which is characterized in that the related data of the waybill to be predicted Carry out data processing, comprising:

Obtain the site code information in the related data of the waybill to be predicted；

The corresponding encoded radio of the site code information is searched in the average Code Mapping Tables constructed in advance.

13. a kind of waybill complaint handling device, which is characterized in that the device includes:

First acquisition unit, for obtaining the related data of waybill to be predicted, the related data of the waybill to be predicted belongs to non- I signs for type；

First data processing unit carries out data processing for the related data to the waybill to be predicted；

Probability prediction unit complains model for the result after data processing to be input to the waybill that constructs in advance, described in output The complaint probability value of waybill to be predicted, the waybill, which complains model, to be obtained using machine learning algorithm training study；And

Developing Tactics unit adjusts the fortune to be predicted for the comparison result based on the complaint probability value and preset threshold value Single corresponding service strategy.

14. device according to claim 13, which is characterized in that the waybill complains the model to include:

First model subelement, for obtaining the first machine learning model using the training study of the first machine learning algorithm；

Second model subelement, for obtaining the second machine learning model using the training study of the second machine learning algorithm；

Subelement is combined, for combining first machine learning model and second machine learning model according to weight Model is complained to waybill.

15. device according to claim 14, which is characterized in that the waybill complains model further include:

Second obtains subelement, and for obtaining history waybill data, the history waybill data belong to non-signs for type in person；

Second data processing subelement, for carrying out data processing to the history waybill data；

Ratio cut partition subelement, for the result after data processing to be divided into training dataset and test number according to preset ratio According to collection.

16. device according to claim 15, which is characterized in that the first model subelement, comprising:

First training submodule, for obtaining first according to the training study of the first machine learning algorithm using the training dataset Machine learning model, first machine learning model are that the characteristic parameter that first machine learning is extracted is signed in person with non- Waybill is complained the mapping relations between probability value, and the characteristic parameter includes at least waybill dimension, user's dimension, part is sent to be tieed up Degree, wherein user's dimension includes user's community information, and described to send part dimension include the corresponding coding of site code information Value；

First test submodule exports the first probability value for the test data set to be inputted the first machine learning model.

17. device according to claim 16, which is characterized in that the second model subelement, comprising:

Second training submodule, for obtaining second according to the training study of the second machine learning algorithm using the training dataset Machine learning model, second machine learning model are waybill dimension, user's dimension, part dimension are sent to sign for waybill in person with non- Mapping relations between complained probability value, user's dimension includes user's community information, and described to send part dimension include site The corresponding encoded radio of code information；

Second test submodule exports the second probability value for the test data set to be inputted the second machine learning model.

18. device according to claim 17, which is characterized in that the combination subelement is used for first probability Value sums to obtain the waybill throwing multiplied by the difference of numerical value 1 and the weight coefficient multiplied by weight coefficient and second probability value Tell model.

19. device described in any one of 3-18 according to claim 1, which is characterized in that the Developing Tactics unit, comprising:

Comparing subunit, for being compared using the complaint probability value with preset threshold value；

Strategy push subelement, is used for if it is greater than the threshold value, based on whether in the presence of the dependency number with the waybill to be predicted Corresponding service strategy is pushed according to associated push recording mark.

20. device according to claim 19, which is characterized in that the strategy, which pushes subelement, to include:

Third obtains module, the first identifier information in related data for obtaining the waybill to be predicted；

Searching module, for being based on first identifier information searching history flag column relevant to the first identifier information Table；

Second judgment module, for judging in the history label list with the presence or absence of push recording mark；

Tactful pushing module, for pushing corresponding service strategy based on the judging result of the push recording mark.

21. device according to claim 20, which is characterized in that the strategy pushing module further include:

First push submodule, for pushing first service strategy if there is no the push recording mark.

22. device according to claim 21, which is characterized in that after the push first service strategy, the device Further include:

Feedback processing modules, it is opposite with addressee information/sender's information in the related data of the waybill to be measured for monitoring The feedback information answered, and second identifier information is generated according to the feedback information.

23. device according to claim 20, which is characterized in that the strategy provides module further include:

4th acquisition submodule, for obtaining in the related data with waybill to be predicted if there is the push recording mark The corresponding second identifier information of addressee information/sender's information, the second identifier information be used to indicate whether exist with The corresponding feedback information of the push recording mark；And

24. the described in any item devices of 3-23 according to claim 1, which is characterized in that the data processing unit, comprising:

5th obtains subelement, the site code information in related data for obtaining the waybill to be predicted；

Subelement is searched, for searching the corresponding volume of the site code information in the average Code Mapping Tables constructed in advance Code value.

25. a kind of computer equipment, can run on a memory and on a processor including memory, processor and storage Computer program, which is characterized in that the processor is realized as described in any in claim 1-12 when executing described program Method.

26. a kind of computer readable storage medium is stored thereon with computer program, the computer program is used for:

The method as described in any in claim 1-12 is realized when the computer program is executed by processor.