CN112163962A

CN112163962A - Method and device for model training and business wind control

Info

Publication number: CN112163962A
Application number: CN202010945381.6A
Authority: CN
Inventors: 王颖; 白亮; 赵翔宇
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2020-09-10
Filing date: 2020-09-10
Publication date: 2021-01-01

Abstract

The specification discloses a method and a device for model training and business wind control. In this specification, a service platform determines, according to acquired historical first service data and user attribute data of a user, each basic feature and a plurality of additional features serving as a wind control feature, then determines, for each wind control feature, a wind control data interval corresponding to the wind control feature, inputs, into a wind control model to be trained, feature data located in the wind control data interval corresponding to each wind control feature, determined from the historical first service data and the user attribute data of the user, trains to obtain the wind control model, and then uses the trained wind control model to perform service wind control on a second service. Therefore, in the embodiment of the description, the insurance business serving as the second business can be subjected to business wind control through the wind control model obtained through training, so that the auditing workload of workers is reduced, and the business execution efficiency is improved.

Description

Method and device for model training and business wind control

Technical Field

The specification relates to the technical field of internet, in particular to a method and a device for model training and business wind control.

Background

With the development of social economy and the improvement of the living standard of people, the insurance business is more and more valued by the public. Conventionally, when a client purchases insurance online, an insurance company usually adopts a manual checking mode to check the health condition of an insurance client to determine whether the insurance client meets insurance application conditions.

At present, the online insurance business is rapidly expanded because the online insurance business is very convenient to execute. At present, insurance companies still adopt the mode of artifical inspection to examine insurable customer's health condition, like this, not only greatly increased the examining and verifying work load of salesmen, the operating efficiency is low moreover.

Disclosure of Invention

The present specification provides a method and apparatus for model training and business scheduling to partially solve the above problems in the prior art.

The technical scheme adopted by the specification is as follows:

the present specification provides a method of model training, comprising:

acquiring historical first service data and user attribute data of each user, wherein the historical first service data is generated by the first service executed historically by the user;

determining each basic characteristic according to the historical first service data and the user attribute data;

determining at least one additional feature from the base features;

taking each basic feature and the at least one additional feature as wind control features, and determining a wind control data interval corresponding to each wind control feature;

for each user, inputting the characteristic data, which are determined from the historical first service data and the user attribute data of the user and are positioned in the wind control data interval corresponding to each wind control characteristic, into a wind control model to be trained to obtain an output result;

and training the wind control model by taking the minimized deviation between the output result and the wind control label corresponding to the user as an optimization target, and carrying out service wind control on the second service through the trained wind control model.

Optionally, the user attribute data includes: at least one of credit data of the user, insurance data of the user, medical data of the user.

Optionally, the historical first traffic data includes: the business address comprises historical order data of the hospital address;

said determining at least one additional feature from said base features comprises:

for each user, determining the order number of historical orders of which the service addresses comprise hospital addresses from historical first service data of the user as a target order number corresponding to the user;

dividing a plurality of order quantity intervals according to the target order quantity corresponding to each user;

and determining additional features according to the order quantity intervals and feature data corresponding to the basic features determined from the user attribute data.

Optionally, the determining additional features according to the order quantity intervals and feature data corresponding to the basic features determined from the user attribute data includes:

clustering the users according to the order quantity intervals and the characteristic data corresponding to the basic characteristics determined from the user attribute data to obtain a plurality of clustering clusters;

determining additional features from the plurality of clusters.

Optionally, the determining, by taking the basic features and the at least one additional feature as wind control features, a wind control data interval corresponding to each wind control feature includes:

for each wind control feature, determining the correlation of the wind control feature on different data intervals, and for each data interval corresponding to the wind control feature, determining an IV corresponding to the wind control feature in the data interval according to the number of users of regular users and the number of users of reverse users, of which the feature data corresponding to the wind control feature is located in the data interval;

and determining a wind control data interval corresponding to the wind control characteristic according to the correlation of the wind control characteristic on different data intervals and the IV corresponding to the wind control characteristic on different data intervals.

Optionally, the wind control model comprises: the system comprises a first wind control model and a second wind control model, wherein the first wind control model is used for carrying out service wind control in a first service stage of a second service, and the second wind control model is used for carrying out service wind control in a second service stage of the second service.

Optionally, if the first wind control model is trained, the obtaining of the historical service data and the user attribute data of each user includes:

taking each user matched with the first service stage as a first user, and acquiring historical first service data and user attribute data of each first user;

determining each basic feature according to the historical first service data and the user attribute data, including:

determining each basic feature aiming at the first service stage as a first basic feature according to historical first service data and user attribute data of each first user;

determining at least one additional feature for the first service phase as a first additional feature according to the first basic features;

the determining, by taking each basic feature and the at least one additional feature as wind control features, a wind control data interval corresponding to each wind control feature includes:

taking each first basic feature and the at least one first additional feature as first wind control features, and determining a wind control data interval corresponding to each first wind control feature;

for each user, inputting the feature data, which is determined from the historical first service data and the user attribute data of the user and is located in the wind control data interval corresponding to each wind control feature, into the wind control model to be trained to obtain an output result, wherein the output result comprises:

for each first user, inputting feature data, which are determined from historical first service data and user attribute data of the first user and are located in a wind control data interval corresponding to each first wind control feature, into a first wind control model to be trained to obtain a first output result;

the training of the wind control model with the minimized deviation between the output result and the wind control label corresponding to the user as an optimization target and the service wind control of the second service through the trained wind control model comprise:

and training the first wind control model by taking the minimum deviation between the first output result and a first wind control label corresponding to the first user in the first service stage as an optimization target, and performing service wind control on the first service stage of the second service through the trained first wind control model.

Optionally, if the second wind control model is trained, the obtaining of the historical service data and the user attribute data of each user includes:

taking each user matched with the second service stage as a second user, and acquiring historical first service data and user attribute data of each second user;

determining each basic feature aiming at the second service stage as a second basic feature according to the historical first service data, the historical second service data and the user attribute data of each second user, wherein the historical second service data is generated by the second user executing a second service historically;

determining at least one additional feature for the second service phase as a second additional feature according to the second basic features;

and taking the second basic features and the at least one second additional feature as second wind control features, and determining a wind control data interval corresponding to each second wind control feature.

Optionally, before the step of inputting, for each user, the feature data, which is determined from the historical first service data and the user attribute data of the user and is located in the wind control data interval corresponding to each wind control feature, into the wind control model to be trained to obtain an output result, the method further includes:

for each second user, inputting feature data, which are determined from historical first service data and user attribute data of the second user and are located in a wind control data interval corresponding to each first wind control feature, into the trained first wind control model to obtain a first wind control result corresponding to the second user;

for each user, inputting feature data corresponding to the wind control features determined from historical first service data and user attribute data of the user into a wind control model to be trained to obtain an output result, wherein the output result comprises:

for each second user, inputting a first wind control result corresponding to the second user and feature data, which are determined from historical first service data, historical second service data and user attribute data of the second user and are located in a wind control data interval corresponding to each second wind control feature, into a second wind control model to be trained to obtain a second output result;

and training the second wind control model by taking the minimum deviation between the second output result and a second wind control label corresponding to the second user in the second service stage as an optimization target, and performing service wind control on the second service stage of the second service through the trained second wind control model.

Optionally, the first service phase of the second service includes: and a second service stage of the second service comprises: and a claim settlement auditing stage.

The present specification provides a method for service wind control, including:

receiving a service request of a user;

according to the service request, determining the corresponding characteristic data of the user in a wind control data interval corresponding to the preset wind control characteristics;

inputting the characteristic data into a pre-trained wind control model to obtain a wind control result, wherein the wind control model is obtained by training through the model training method;

and carrying out service wind control on the user according to the wind control result.

The present specification provides an apparatus for model training, comprising:

the data acquisition module is used for acquiring historical first service data and user attribute data of each user, wherein the historical first service data is generated by the first service executed by the user in history;

a basic feature determination module, configured to determine each basic feature according to the historical first service data and the user attribute data;

an additional feature determination module for determining at least one additional feature based on the base features;

the wind control data interval determining module is used for taking each basic feature and the at least one additional feature as wind control features and determining a wind control data interval corresponding to each wind control feature according to each wind control feature;

the output module is used for inputting the characteristic data which are determined from the historical first service data and the user attribute data of each user and are positioned in the wind control data interval corresponding to each wind control characteristic into the wind control model to be trained to obtain an output result;

and the training module is used for training the wind control model by taking the deviation between the minimized output result and the wind control label corresponding to the user as an optimization target, and carrying out service wind control on the second service through the trained wind control model.

The device of business wind accuse that this specification provided includes:

the receiving module is used for receiving a service request of a user;

the characteristic data determining module is used for determining the characteristic data corresponding to the user in a wind control data interval corresponding to the preset wind control characteristic according to the service request;

the wind control result determining module is used for inputting the characteristic data into a pre-trained wind control model to obtain a wind control result, and the wind control model is obtained by training through the model training method;

and the service wind control module is used for carrying out service wind control on the user according to the wind control result.

The present specification provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described method of model training or business scheduling.

The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the model training method or the business wind control method when executing the program.

The technical scheme adopted by the specification can achieve the following beneficial effects:

in the method for model training and business wind control provided by this specification, a business platform determines each basic feature and a plurality of additional features as wind control features according to acquired historical first business data and user attribute data of a user, then determines a wind control data interval corresponding to each wind control feature for each wind control feature, and inputs feature data, determined from the historical first business data and user attribute data of the user, located in the wind control data interval corresponding to each wind control feature into a wind control model to be trained, the wind control model is obtained by training, and then the trained wind control model is used to perform business wind control on a second business. Therefore, in the embodiment of the description, the insurance business serving as the second business can be subjected to business wind control through the wind control model obtained through training, so that the auditing workload of workers is reduced, and the business execution efficiency is improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the specification and are incorporated in and constitute a part of this specification, illustrate embodiments of the specification and together with the description serve to explain the specification and not to limit the specification in a non-limiting sense. In the drawings:

FIG. 1 is a schematic flow chart of a method of model training in the present specification;

fig. 2 is a schematic flow chart of a method for service wind control in this specification;

FIG. 3 is a schematic flow chart of a method for model training of a wind control model corresponding to an underwriting stage in this specification;

fig. 4 is a schematic flow chart of a method for service wind control of a wind control model corresponding to an insurance application auditing stage in this specification;

fig. 5 is a schematic flowchart of a method for model training of a wind-controlled model corresponding to a claim auditing stage in this specification;

fig. 6 is a schematic flow chart of a method for service wind control of a wind control model corresponding to a claim auditing stage in this specification;

FIG. 7 is a schematic diagram of an apparatus for model training provided herein;

fig. 8 is a schematic diagram of an apparatus for traffic throttling provided herein;

fig. 9 is a schematic diagram of an electronic device corresponding to the model training method shown in fig. 1 or the traffic wind control method shown in fig. 2 provided in the present specification.

Detailed Description

In order to make the objects, technical solutions and advantages of the present disclosure more clear, the technical solutions of the present disclosure will be clearly and completely described below with reference to the specific embodiments of the present disclosure and the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort belong to the protection scope of the present specification.

The embodiment of the specification can be used for risk control of insurance business, wherein the insurance business comprises two business phases, namely an insurance application auditing phase and a claim settlement auditing phase.

In the embodiment of the present specification, two service phases of the insurance service respectively correspond to one wind control model, and the wind control model includes two sub-models, one is a healthy wind control model, and the other is a fraud wind control model.

As a whole, the training process of each sub-model in the embodiments of the present specification is substantially the same, and the differences are only: the feature dimensions of the training data of the selected input model may be different due to different stages of model application and different targets of model training.

Therefore, in the embodiments of the present specification, the stage of model application and the target of model training are not considered, and only a specific process of model training is briefly described.

The technical solutions provided by the embodiments of the present description are described in detail below with reference to the accompanying drawings.

Fig. 1 is a schematic flow chart of a model training method in this specification, which specifically includes the following steps:

step S100, obtaining historical first service data and user attribute data of each user, where the historical first service data is generated when the user executes a first service historically.

In insurance business, two phases are typically involved: the stage of insuring the applicant and the stage of applying for claim. In practical application, when an insurance applicant purchases insurance from an insurance company, an insurance company carries out qualification audit on the insurance applicant before formally confirming that the insurance applicant applies insurance, so as to confirm whether an insured person applying insurance by the insurance applicant meets insurance application conditions or not, and refuses the insurance purchasing requirement of the insurance applicant when the insured person does not meet the insurance application conditions, so as to reduce business risks. For example, when an insured person buys a large risk for himself, the insurance company needs to investigate the health condition of the insured person to avoid the situation that the insured person cheats on obtaining insurance funds with the disease.

When an applicant applies for a claim, an insurance company checks the reason of the claim submitted by the applicant, and judges whether the claim application of the applicant is true or not and whether the claim application conforms to the claim condition in an issued insurance contract or not so as to determine whether to carry out claim payment aiming at the claim application of the applicant and reduce the business risk. For example, after an applicant buys a large risk for himself, the applicant is hospitalized during the period when the risk is effective and applies for insurance claims to an insurance company who buys the risk, after the insurance company receives the claim application of the applicant, the insurance company needs to investigate the actual situation of the claim application of the applicant, determine whether the situation of the claim application of the applicant is substantial, and whether the claim application of the applicant meets the claim condition, if so, pay the claim for the claim application of the applicant.

Based on this, in the embodiment of the present specification, historical first service data and user attribute data of each user are used as sample data of model training, a wind control model of the insurance service is obtained through training, and the wind control model is used to perform risk control on the service of the insurance applicant, so as to reduce the workload of service personnel and improve the service execution efficiency of the insurance service.

The historical first service data of the user is service data generated by the first service historically executed by the user. The first business refers to other businesses which have been executed by the user historically except for insurance businesses which use trained wind control models for risk control. For example, the order details of the user purchasing takeaway on the takeaway platform, the order details of the user purchasing medicine on the e-commerce platform, the life service payment record on the online platform, and so on can all be referred to as the historical first business data of the user.

In the embodiment of the present specification, the user attribute data is used to reflect some practical situations of the user, for example, the user attribute data may include: credit data for the user, insurance data for the user, medical data for the user, and the like.

The credit data of the user refers to all data which can reflect the credit state of the user. For example, the credit data of the user may include the credit rating of the user, credit information of the user, credit records of the user, default records of the user, and the like.

The insurance data of the user refers to the insurance data of the insurance business purchased by the user, wherein the insurance business does not include the insurance business which uses the trained wind control model for risk control in the embodiment of the specification. For example: the insurance data of the user may include whether the user has an insurance service purchased, service data of the insurance service purchased by the user, and the like.

The medical data of the user may include a user appointment registry record, historical visit records, illness records, and the like.

Of course, in addition to the above-mentioned several user attribute data, the user attribute data may also include basic attribute information of the user, such as user age, user gender, user native place, user occupation, user residential address, and so on, which will not be described in detail herein.

As can be seen from the above, the acquired historical first service data and the user attribute data often include data of a plurality of characteristic dimensions. Therefore, after the historical first service data and the user attribute data are acquired, the acquired historical first service data and the acquired user attribute data need to be analyzed and processed, and the basic features suitable for the wind control model are determined.

In the embodiment of the present specification, the execution main body for training the wind control model and executing the service wind control in the subsequent process may be a terminal device such as a computer, or may be a service platform composed of a server, a terminal, and the like, and the service platform may provide an insurance service to a user. For convenience of description, the model training method and the business wind control method in the embodiment of the present specification will be described below by taking only the business platform as an execution subject as an example.

Step S102, determining each basic characteristic according to the historical first service data and the user attribute data.

In step S102, after the historical first service data and the user attribute data used for training the model are obtained, the obtained data need to be analyzed, and the basic features applicable to the wind control model to be trained are determined.

For example, the first service is an online shopping service, the acquired historical first service data of the user includes an order detail of online shopping of the user in one year, and the characteristic dimensions that can be counted from the acquired order detail include: the total amount of orders of online shopping of users in one year, the list of commodities in the orders of online shopping of users in one year, the category of the commodities in the orders of online shopping of users in one year, the number of times that the addresses of the orders of online shopping of users in one year are hospitals, the total consumption amount of the orders of online shopping of users in one year and the like.

If the online shopping order is a take-away order, the determined basic characteristics may include: the delivery address of the takeout order of the user within one year is the number of times of the hospital, the food type of the takeout order online purchased by the user within one year, and the like.

If the online shopping order is a medicine order, the basic characteristics can be determined from the medicine order, and the basic characteristics comprise: the name of the medicine purchased by the user on line within one year, the disease treated by the medicine purchased by the user on line within one year, the department to which the disease treated by the medicine purchased by the user on line within one year belongs, the adverse reaction corresponding to each medicine purchased by the user on line within one year and the like.

Of course, the service platform may also determine the basic features from the obtained user attribute data. For example, the basic features that the service platform may determine from the credit data of the user may include: whether each loan item currently established by the user is due, the user's credit, the user's loan amount, etc.

For the insurance data of the user, the basic characteristics that the service platform can determine from include: whether the user has the insurance service purchased, the purchase amount of the insurance service purchased by the user, the guarantee items of the insurance service purchased by the user, the insurance amount of the insurance service purchased by the user, and the like.

For the medical data of the user, the basic features that the service platform can determine from include: the number of times the user makes a appointment, the hospital the user makes a appointment, the department the user makes a appointment, etc.

The above example is only for better explaining how to determine the basic features based on the acquired data after acquiring the historical first service data and the user attribute data. The basic characteristics determined according to the actual business requirements are not exactly the same, so the above examples are only used for illustration and do not limit the scope of the present invention.

And step S104, determining at least one additional characteristic according to the basic characteristics.

In step S104, in order to further enrich the wind control dimensions corresponding to the wind control model, some additional features may be determined according to the determined basic features. Wherein, the basic characteristic for determining the additional characteristic can be selected according to the requirement.

In embodiments of the present description, the specific manner of determining the additional features may be varied. For example: when the basic features include the number of times that the address of the order purchased by the user in the online manner is the number of times of the hospital in one year, the basic features can be divided into intervals according to the feature data corresponding to the basic features, and the divided intervals of the number of times can be used as additional features.

The mode of dividing the regions can be various, for example, the regions are divided manually into a number of times, users whose receiving addresses are hospital orders with the number of 0-5 are divided into a first time region, and users whose receiving addresses are hospital orders with the number of 6-10 are divided into a second time region; … … are provided. Therefore, the address of the order for the online purchase of the user in one year is the characteristic of the number of times of the hospital, and the address can be converted into a plurality of times intervals. For another example, each user may be clustered by using a method such as K-means clustering according to the number of times that the address of the order online purchased by the user is a hospital in one year, so as to obtain a plurality of cluster clusters, and the cluster clusters are used as additional features.

Of course, the service platform may also determine the additional features by combining a plurality of basic features. As a possible implementation, the service platform may select one basic feature from the basic features determined according to the historical first service data and one basic feature from the basic features determined according to the user attribute data, and then determine an additional feature according to the two selected basic features.

The service platform divides the feature data corresponding to one of the basic features into a plurality of feature data intervals, and determines the additional features according to the divided feature data intervals and the feature data of the other basic feature. Specifically, if the basic feature determined from the historical first service data is the order number of the historical order whose service address includes the hospital address, for each user, the order number of the historical order whose service address includes the hospital address is determined from the historical first service data of the user, and is used as the target order number corresponding to the user, a plurality of order number intervals are divided according to the target order number corresponding to each user, and the additional feature is determined according to the plurality of order number intervals and the feature data corresponding to the basic feature determined from the user attribute data. When the business platform determines that the business address contains the order number of the historical orders of the hospital address, the first historical business order of the user needs to be obtained, each receiving address is matched with the order of the hospital in a fuzzy mode in the business order detail, and the order number of the hospital in the receiving address is counted.

The basic characteristics determined by the service platform from the user attribute data can be the user age, the user gender, the user native place and the like. Then, the service platform can cluster each user according to the divided order quantity intervals and the characteristic data corresponding to the basic characteristics determined from the user attribute data to obtain a plurality of cluster clusters, and determine the additional characteristics according to the cluster clusters.

Taking the example that the basic feature determined from the user attribute data of the user is the age of the user, the service platform can cluster the users according to the divided order quantity intervals and the ages of the users to obtain a plurality of cluster clusters, and each cluster can be used as a feature interval of the additional feature.

And step S106, taking each basic feature and the at least one additional feature as wind control features, and determining a wind control data interval corresponding to each wind control feature.

In practical application, because the data volume of the acquired historical first service data and the acquired user attribute data is large, the dimensionalities of the correspondingly determined basic features and the dimensionalities of the corresponding extra features are large, and if the feature data corresponding to all the wind control features are input to a model needing to be trained without screening, the situations that the model training fitting speed is slow, the stability of the trained model is poor and the like may occur.

Therefore, in the embodiment of the present specification, the service platform may use all basic features and all additional features as the wind control features, determine a wind control data interval corresponding to the wind control features, and input data of the user located in the wind control data interval into the wind control model for training.

In step S106, the service platform first divides each wind control feature into a plurality of different data intervals, then determines, for each wind control feature, a correlation of the wind control feature in the different data intervals, and simultaneously determines, for each data interval corresponding to the wind control feature, an Information Value (IV) corresponding to the wind control feature in the data interval according to the number of users of the regular users and the number of users of the reverse users, where the feature data corresponding to the wind control feature is located in the data interval, and finally determines, according to the correlation of the wind control feature in the different data intervals and the IV corresponding to the wind control feature in the different data intervals, the wind control data interval corresponding to the wind control feature.

The service platform may determine the correlation of the wind control characteristic in different data intervals by a method such as Principal Component Analysis (PCA). Of course, besides using PCA to determine the correlations between different data intervals of each of the wind-control features, the service platform may also use other methods to determine each set of correlation features, such as Singular Value Decomposition (SVD).

Further, since each user is labeled in advance, the service platform can directly determine whether each user is a regular user (the regular user is a user who has determined that the user does not have a risk of being controlled by the wind, i.e., a security user) or a counterexample user (the counterexample user is a user who has determined that the user has a risk of being controlled by the wind, i.e., a risk user). Thus, when determining the IV corresponding to the wind control feature in the wind control data interval for each wind control feature, the service platform may determine, according to the number of users of the positive-case users and the number of users of the negative-case users located in the wind control data interval, that the IV corresponding to the wind control feature in the wind control data interval needs to be described. The specific IV calculation formula is as follows:

wherein H1 is the number of positive users in a data interval, H1T is the total number of positive users, H2 is the number of negative users in a data interval, H1T is the total number of negative users, and n is the number of user classes after classifying the users, that is, the number of data intervals. The service platform may partition a plurality of data intervals for the wind control characteristic based on the above formula, for example, it may be enough to ensure that an IV value determined based on the partitioned data intervals meets a preset condition.

Further, the IV of the wind control characteristic in each data interval can be actually determined through the above formula, so that the wind control data interval corresponding to the wind control characteristic can be determined by combining the determined correlations of the wind control characteristic in different data intervals.

For example, if it is assumed that after a plurality of age groups are divided for the wind control feature of the user age, in these age groups, there are two age groups 1 and 2 having a high correlation, it is necessary to identify the IV corresponding to the wind control feature in each of the two age groups, and if it is determined that the IV corresponding to the wind control feature in the age group 1 is larger than the IV corresponding to the wind control feature in the age group 2, the age group 1 may be used as the wind control data group corresponding to the wind control feature.

It should be noted that, in the embodiment of the present specification, a wind control data interval corresponding to one wind control feature does not have only one data interval, and actually, one wind control feature may correspond to a plurality of wind control data intervals, in these wind control data intervals, some wind control data intervals are screened by determining the IV, and some wind control data intervals belong to data intervals with low correlation with other data intervals.

Of course, for each wind control feature, the service platform may also use all data intervals corresponding to the wind control feature as wind control data intervals, and may set a corresponding weight value for each wind control data interval to reduce the influence of the correlation between features on the model stability. Specifically, the size of the IV corresponding to the wind control data interval of the wind control characteristic and the size of the weight value corresponding to each wind control data interval are in a positive correlation relationship, if the IV corresponding to the wind control data interval is larger, the weight value corresponding to the wind control data interval is also larger, and if not, the weight value is smaller. Therefore, when the model is trained, more reference data are provided for the model in the wind control data interval with the large weight value, and the situation that the stability of the trained wind control model is poor due to the correlation among the features can be avoided.

And S108, inputting the characteristic data, which are determined from the historical first service data and the user attribute data of each user and are positioned in the wind control data interval corresponding to each wind control characteristic, into the wind control model to be trained to obtain an output result.

And step S110, training the wind control model by taking the deviation between the minimized output result and the wind control label corresponding to the user as an optimization target, and carrying out service wind control on the second service through the trained wind control model.

In the embodiment of the present specification, after the service platform determines each wind control feature and the wind control data interval corresponding to each wind control feature, the feature data, which is determined from the historical first service data of the user and the user attribute data and is located in the wind control data interval corresponding to each wind control feature, may be input into the wind control model to be trained to obtain a corresponding output result, and then, a supervised training mode may be adopted to train the wind control model with a goal of minimizing a deviation between the output result and the wind control label corresponding to the user as an optimization goal, so that in a subsequent process, the trained wind control model performs service wind control on a second service executed by the user.

Therefore, in the embodiment of the description, the insurance business serving as the second business can be subjected to business wind control through the wind control model obtained through training, so that the auditing workload of workers is reduced, and the business execution efficiency is improved.

For the wind control model trained by the model training method, the embodiment of the specification further provides a using method of the wind control model.

Fig. 2 is a schematic flow chart of a method for controlling a service profile according to an embodiment of the present disclosure, which specifically includes the following steps:

step 200, receiving a service request of a user.

The service request may be an insurance Application request submitted to the service platform when the user purchases insurance on a terminal device (such as a mobile phone, a tablet computer, etc.) or software such as a client and an Application (App) installed on the terminal device, or a claim settlement Application request submitted to the service platform for the purchased insurance on the terminal device or software such as the client and the App installed on the terminal device.

Step 202, according to the service request, determining the feature data corresponding to the user in the wind control data interval corresponding to the preset wind control feature.

In the step, the service platform firstly determines a current service stage according to a received service request of a user, and then determines an adopted wind control model and characteristic data corresponding to the user in a wind control data interval corresponding to a preset wind control model according to the determined service stage.

The wind control data interval corresponding to the preset wind control characteristics is the wind control data interval in which data needing to be input into the wind control model is located when the wind control model is determined based on the service phase of the model during model training. The characteristic data corresponding to the user in the wind control data interval corresponding to the preset wind control characteristic refers to characteristic data located in the wind control data interval corresponding to each wind control characteristic and determined from historical first service data and user attribute data of the user.

For example, after a plurality of age intervals are divided according to the age of the user, two age intervals corresponding to the wind control feature of the age of the user are provided: the age section 1 and the age section 2 are respectively provided, wherein the characteristic data range corresponding to the age section 1 is 30 years to 39 years, and the characteristic data range corresponding to the age section 2 is 40 years to 49 years. If the service platform determines that the age of the user is 35 years after receiving the application request of the user for insuring, and the service platform is located in the feature data range corresponding to the age interval 1, the feature data of the age of 35 years is the feature data corresponding to the user in the wind control data interval corresponding to the preset wind control feature. If the service platform determines that the age of the user is 25 years after receiving the application request of the user for insuring, and the user is not included in the feature data range corresponding to the age interval 1 or the age interval 2, the feature data of the age of 25 is not the feature data corresponding to the user in the wind control data interval corresponding to the preset wind control feature.

And 204, inputting the characteristic data into a pre-trained wind control model to obtain a wind control result.

And step 206, performing service wind control on the user according to the wind control result.

In the embodiment of the specification, a service platform receives a service request of a user, and determines corresponding feature data of the user in a wind control data interval corresponding to preset wind control features according to the service request; and inputting the characteristic data into a pre-trained wind control model to obtain a wind control result, and carrying out service wind control on the user according to the wind control result. When the second service is an insurance service, the service wind control can be performed on the insurance service through the wind control model obtained by training in the embodiment of the specification. Therefore, in the embodiment of the description, the insurance business serving as the second business can be subjected to business wind control through the wind control model obtained through training, so that the auditing workload of workers is reduced, and the business execution efficiency is improved.

For the case that the second service is an insurance service, the first service phase of the second service comprises: and a second service stage of the second service comprises: in the claim auditing stage, the wind control model corresponding to the first service stage may be referred to as a first wind control model, and the wind control model corresponding to the second service stage may be referred to as a second wind control model.

Further, the first wind control model trained for the first business stage may include two sub-risk models, which are a health risk sub-model and a fraud risk sub-model, respectively, and the second wind control model trained for the second business stage may also include two sub-risk models, which are a health risk sub-model and a fraud risk sub-model. In the specific training and using process, due to different service phases, the input data and the output result in the first wind control model are not identical to the input data and the output result in the second wind control model.

The training method and the using method of the wind control model corresponding to the two service phases of the insurance service will be described below.

Fig. 3 is a schematic flowchart of a method for model training in an underwriting stage according to an embodiment of the present disclosure, which specifically includes the following steps:

and step S300, taking each user matched with the insurance application auditing stage as a first user, and acquiring historical first service data and user attribute data of each first user.

When the first wind control model corresponding to the insurance application auditing stage is trained, all users matched with the insurance application auditing stage can refer to all users registered on the service platform.

Step S302, determining each basic feature aiming at the insurance application auditing stage as a first basic feature according to the historical first service data and the user attribute data of each first user.

Step S304, determining at least one additional characteristic aiming at the underwriting auditing stage as a first additional characteristic according to the first basic characteristics.

Step S306, using the first basic features and the at least one first additional feature as first wind control features, and determining, for each first wind control feature, a wind control data interval corresponding to the first wind control feature.

Step S308, for each first user, inputting the feature data, which are determined from the historical first service data and the user attribute data of the first user and are located in the wind control data interval corresponding to each first wind control feature, into the first wind control model to be trained to obtain a first output result.

And S310, training the first wind control model by taking the minimum deviation between the first output result and a first wind control label corresponding to the first user in the insurance application auditing stage as an optimization target, and performing service wind control on the second service in the insurance application auditing stage through the trained first wind control model.

As can be seen from the above steps, for the first wind control model corresponding to the first service stage of the second service, the manner adopted when training the first wind control model is basically the same as the manner shown in the above steps S102 to S110. It can be understood that the model training process described in steps S102 to S110 is applicable to the wind control model corresponding to each service phase of the second service.

Fig. 4 is a schematic flowchart of a method for using a wind control model in an underwriting stage according to an embodiment of the present disclosure, which specifically includes the following steps:

step 400, receiving a service request of a user.

Step 402, according to the service request, determining the feature data corresponding to the user in the wind control data interval corresponding to the preset wind control feature.

The method comprises the steps of firstly determining that a current service stage is an insurance application and verification stage (namely a first service stage of a second service) according to a received service request of a user, then determining a first wind control model adopted in the insurance application and verification stage and wind control characteristics corresponding to the first wind control model, and determining characteristic data corresponding to the user in a wind control data interval corresponding to preset wind control characteristics.

And step 404, inputting the characteristic data into a pre-trained wind control model to obtain a wind control result.

And 406, performing service wind control on the user according to the wind control result.

If the first wind control model corresponding to the insurance application and verification stage comprises a health risk sub-model and a fraud risk sub-model, the wind control results output by the two models are respectively the health risk level and the fraud risk level of the user.

Specifically, for the case that the first wind control model is a model, when a user applying for insurance uses the first wind control model to perform business wind control, if the wind control result is high risk, the user's request for applying for insurance is intercepted, if the wind control result is medium risk, the user is allowed to join, and a continuous attention mark is marked; and if the wind control result is low risk, the user is allowed to join.

When the first wind control model comprises the health risk sub-model and the fraud risk sub-model, and the two models are used for carrying out business wind control on the user, the wind control results of the two models need to be comprehensively considered. The concrete measures for performing service wind control on the user according to the wind control results output by the two models can be seen in the following table:

table 1, health risk submodel and corresponding relation between wind control result of fraud risk submodel and business wind control measure in insurance application auditing stage

The marking of the user mentioned in table 1 means that the business data of the user is continuously monitored in the following period to further determine whether the user has a potential business risk.

It should be noted that if the first wind control model includes the health risk sub-model and the fraud risk sub-model, the risk dimensions corresponding to the two models are different, that is, in the model training phase, the two models may be trained separately, the data sample sets adopted by the two models may be the same, but in the model training phase, the wind control characteristics and the wind control data interval applicable to the two models are determined separately. Therefore, in the actual use stage, after receiving the service request of the user, the service platform can respectively determine the feature data corresponding to the user in the two models according to the service request, and input the feature data into the two models, and the two models can also respectively output corresponding wind control results.

Fig. 5 is a schematic flowchart of a method for model training at an claim review stage according to an embodiment of the present disclosure, which specifically includes the following steps:

and step S500, taking each user matched with the claim auditing stage as a second user, and acquiring historical first service data and user attribute data of each second user.

And in the claim auditing stage, the second user matched with the claim auditing stage is a user which passes the insurance application auditing stage and successfully joins the insurance application on the service platform.

Step S502, determining, as second basic features, basic features for the claim audit stage according to the historical first service data, the historical second service data, and the user attribute data of each second user, where the historical second service data is generated by the second user performing a second service historically.

Wherein the second service data is service data generated when the user historically executes the second service. For example, where the second service is an insurance service, the historical second service data referred to herein includes insurance data generated by insured users during an insurance review phase and/or a claim settlement review phase. When the second service data includes service data generated in the claim auditing stage of the insured user, the service data mentioned herein may include items, money amount, etc. of the user insurable when performing the second service, and also include reasons and address information for fuzzy matching of the user's claim from the claim application information and the claim proof filled by the user. The address information mentioned here can be the address where the claim application event occurs, or the hospital address for the disease diagnosis of claim application, etc.

Step S504, determining at least one additional feature aiming at the claim auditing stage according to the second basic features, and using the additional feature as a second additional feature;

step S506, using the second basic features and the at least one second additional feature as second wind control features, and determining, for each second wind control feature, a wind control data interval corresponding to the second wind control feature.

Step S508, for each second user, inputting the feature data, which is located in the wind control data interval corresponding to each first wind control feature and is determined from the historical first service data and the user attribute data of the second user, into the trained first wind control model, so as to obtain a first wind control result corresponding to the second user.

Step S510, for each second user, inputting a first wind control result corresponding to the second user and feature data, which is determined from historical first service data, historical second service data, and user attribute data of the second user and is located in a wind control data interval corresponding to each second wind control feature, into a second wind control model to be trained, so as to obtain a second output result;

and step S512, training the second wind control model by taking the minimum deviation between the second output result and a second wind control label corresponding to the second user in the claim auditing stage as an optimization target, and performing service wind control on the claim auditing stage of the second service through the trained second wind control model.

It can be seen from this that, when training the second wind control model corresponding to the second service phase of the second service, it is a result that the first wind control model needs to be output. It is worth emphasizing that, because the output result of the first wind control model may be inaccurate when the training is not completed, it is required to train the first wind control model first, and after the training is completed, the trained first wind control model is required to output the corresponding wind control result, i.e. the first wind control result, to each user.

When the second wind control model is trained, besides the feature data which are determined from the historical first service data, the historical second service data and the user attribute data of the second user and are located in the wind control data interval corresponding to the second wind control feature are required to be input into the second wind control model to be trained, the first wind control result is required to be input into the second wind control model. That is, in the embodiment of the present specification, the result output by the second wind control model refers to the result output by the first wind control model.

Fig. 6 is a schematic flowchart of a method used by a wind control model in a claim audit stage according to an embodiment of the present specification, and specifically includes the following steps:

step 600, receiving a service request of a user.

The service request is a claim settlement request submitted when a user applies for claim settlement on terminal equipment or client, App and other software.

Step 602, according to the service request, determining the feature data corresponding to the user in the wind control data interval corresponding to the preset wind control feature.

In the claim auditing stage, the characteristic data comprise characteristic data which are extracted from historical first service data, historical second service data and user attribute data of the user and are positioned in a wind control data interval corresponding to second wind control characteristics, and in the insurance approving stage, the user and the user output wind control results aiming at the user through a wind control model corresponding to the insurance approving stage by the service platform.

And step 604, inputting the characteristic data into a pre-trained wind control model to obtain a wind control result.

And 606, performing service wind control on the user according to the wind control result.

If the second wind control model corresponding to the claim auditing stage includes a health risk sub-model and a fraud risk sub-model, the wind control results output by the two models are respectively the health risk level and the fraud risk level of the user.

Specifically, for the case that the second wind control model is one model, when the claim application user uses the second wind control model to perform business wind control, if the wind control result is high risk or medium risk, the key attention mark is marked, and if the wind control result is low risk, the claim settlement process can be directly executed.

In the process, if the second service relates to disease risk, when the key attention mark is marked, the disease which the user may suffer from can be estimated according to the historical first service data, the historical second service data and the user attribute data of the user.

For the case that the second wind control model comprises the health risk sub-model and the fraud risk sub-model, when the two models are used for carrying out business wind control on the user, the wind control results of the two models need to be comprehensively considered. How to execute claims according to the wind control results output by the two models can be determined according to actual business requirements, and is not illustrated in detail here. It should be noted that, if the second wind control model includes the health risk sub-model and the fraud risk sub-model, the risk dimensions corresponding to the two models are different. That is to say, in the model training stage, the two models may be trained separately, and the data sample sets adopted by the two models may be the same, but in the model training stage, the wind control features and the wind control data intervals applicable to the two models need to be determined separately. Therefore, in the actual use stage, after receiving the service request of the user, the service platform can respectively determine the feature data corresponding to the user in the two models according to the service request, and input the feature data into the two models, and the two models can also respectively output corresponding wind control results.

Further, when the second wind control model is trained and used, the trained wind control result output by the first wind control model for the user needs to be used as input data. Therefore, if the second wind control model also includes the health risk sub-model and the fraud risk sub-model, the output results of the health risk sub-model and the fraud risk sub-model in the first business stage also need to be input into each sub-model in the second business stage. Taking the health risk sub-model as an example, in the claims auditing stage, the business platform needs to input the wind control result output by the health risk sub-model of the user in the first business stage into the health risk sub-model in the second business stage.

It should be further noted that, in the embodiment of the foregoing description, only the insurance service is taken as an example for description, and actually, the model training method and the service wind control method provided in this description may also be applied to other service scenarios. For example, the credit card business relates to a credit card application phase and a cash withdrawal application phase, in the credit card application phase, a wind control model can be corresponded to identify whether the user has qualification for applying the credit card through the wind control model, and in the cash withdrawal application phase, a wind control model can be corresponded to identify whether the user has qualification for withdrawing the cash and whether the credit amount exceeds the set credit amount through the wind control model.

It should be emphasized that, for different service scenarios, the wind control characteristics corresponding to the wind control model and the wind control data intervals corresponding to the wind control characteristics are different, but the model training method and the service wind control method provided in this specification are applicable to other service scenarios similar to the insurance service, and the specific situations of the other service scenarios are not illustrated in detail here.

Based on the same idea, the present specification further provides a corresponding model training apparatus, as shown in fig. 7.

FIG. 7 provides an apparatus for model training, provided by the present specification for an embodiment of the present specification, including:

a data obtaining module 700, configured to obtain historical first service data and user attribute data of each user, where the historical first service data is service data generated by a user performing a first service historically;

a basic feature determining module 701, configured to determine each basic feature according to the historical first service data and the user attribute data;

an additional feature determination module 702, configured to determine at least one additional feature according to the basic features;

a wind control data interval determining module 703, configured to use each basic feature and the at least one additional feature as a wind control feature, and determine, for each wind control feature, a wind control data interval corresponding to the wind control feature;

an output module 704, configured to, for each user, input feature data, which is determined from historical first service data and user attribute data of the user and is located in a wind control data interval corresponding to each wind control feature, into a wind control model to be trained, so as to obtain an output result;

the training module 705 is configured to train the wind control model with a minimization of a deviation between the output result and the wind control label corresponding to the user as an optimization target, and perform service wind control on the second service through the trained wind control model.

an additional feature determining module 702, configured to determine, for each user, an order number of a historical order whose business address includes a hospital address from historical first business data of the user, as a target order number corresponding to the user; dividing a plurality of order quantity intervals according to the target order quantity corresponding to each user; and determining additional features according to the order quantity intervals and feature data corresponding to the basic features determined from the user attribute data.

Optionally, the additional feature determining module 702 is specifically configured to, when determining an additional feature according to the number of order quantity intervals and feature data corresponding to the basic feature determined from the user attribute data, cluster the users according to the number of order quantity intervals and feature data corresponding to the basic feature determined from the user attribute data to obtain a plurality of cluster clusters; determining additional features from the plurality of clusters.

A wind control data interval determining module 703, configured to determine, for each wind control feature, a correlation of the wind control feature in different data intervals, and determine, for each data interval corresponding to the wind control feature, an IV corresponding to the wind control feature in the data interval according to the number of users of positive-case users and the number of users of negative-case users, of feature data corresponding to the wind control feature, which are located in the data interval; and determining a wind control data interval corresponding to the wind control characteristic according to the correlation of the wind control characteristic on different data intervals and the IV corresponding to the wind control characteristic on different data intervals.

Optionally, if the first wind control model is trained,

a data obtaining module 700, configured to use each user matched with the first service stage as a first user, and obtain historical first service data and user attribute data of each first user;

a basic feature determining module 701, configured to determine, according to the historical first service data and the user attribute data of each first user, each basic feature for the first service stage as a first basic feature;

an additional feature determining module 702, configured to determine, according to the first basic features, at least one additional feature for the first service phase as a first additional feature;

a wind control data interval determining module 703, configured to specifically use each of the first basic features and the at least one first additional feature as a first wind control feature, and determine, for each first wind control feature, a wind control data interval corresponding to the first wind control feature;

an output module 704, specifically configured to, for each first user, input feature data, which is determined from historical first service data and user attribute data of the first user and is located in a wind control data interval corresponding to each first wind control feature, into a first wind control model to be trained, so as to obtain a first output result;

the training module 705 is specifically configured to train the first wind control model with a minimization of a deviation between the first output result and a first wind control tag corresponding to the first user in the first service phase as an optimization target, and perform service wind control on the first service phase of the second service through the trained first wind control model.

Optionally, if the second wind control model is trained,

a data obtaining module 700, configured to specifically use each user matched with the second service phase as a second user, and obtain historical first service data and user attribute data of each second user;

a basic feature determining module 701, configured to determine, according to the historical first service data, the historical second service data, and the user attribute data of each second user, each basic feature for the second service stage as a second basic feature, where the historical second service data is generated by the second user performing a second service historically;

an additional feature determining module 702, configured to determine, according to the second basic features, at least one additional feature for the second service phase as a second additional feature;

the wind control data interval determining module 703 is specifically configured to use each of the second basic features and the at least one second additional feature as a second wind control feature, and determine, for each second wind control feature, a wind control data interval corresponding to the second wind control feature.

Optionally, the apparatus further comprises:

the output module 704 is further configured to, for each user, input the feature data, determined from the historical first service data and the user attribute data of the user, located in the wind control data interval corresponding to each wind control feature into the wind control model to be trained, and before obtaining an output result, for each second user, input the feature data, determined from the historical first service data and the user attribute data of the second user, located in the wind control data interval corresponding to each first wind control feature into the trained first wind control model, so as to obtain a first wind control result corresponding to the second user;

the output module 704 is specifically configured to, for each second user, input a first wind control result corresponding to the second user and feature data, which is determined from historical first service data, historical second service data, and user attribute data of the second user and is located in a wind control data interval corresponding to each second wind control feature, into a second wind control model to be trained, so as to obtain a second output result;

the training module 705 is specifically configured to train the second wind control model with a minimization of a deviation between the second output result and a second wind control label corresponding to the second user in the second service phase as an optimization target, and perform service wind control on the second service phase of the second service through the trained second wind control model.

Based on the same idea, the present specification further provides a corresponding device for service wind control, as shown in fig. 8.

Fig. 8 is a device for service wind control provided in an embodiment of the present specification, including:

a receiving module 800, configured to receive a service request of a user;

a feature data determining module 801, configured to determine, according to the service request, feature data corresponding to the user in a wind control data interval corresponding to a preset wind control feature;

a wind control result determining module 802, configured to input the feature data into a wind control model trained in advance to obtain a wind control result, where the wind control model is obtained by training through a model training method in an embodiment of the present specification;

and the service wind control module 803 is configured to perform service wind control on the user according to the wind control result.

The present specification also provides a computer-readable storage medium having stored thereon a computer program operable to perform the method of model training provided in fig. 1 or the method of traffic scheduling provided in fig. 2 described above.

This specification also provides a schematic block diagram of the electronic device shown in fig. 9. As shown in fig. 9, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, but may also include hardware required for other services. The processor reads a corresponding computer program from the non-volatile memory into the memory and then runs the computer program to implement the method of model training provided in fig. 1 or the method of traffic scheduling provided in fig. 2. Of course, besides the software implementation, the present specification does not exclude other implementations, such as logic devices or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logic unit, and may be hardware or logic devices.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the various elements may be implemented in the same one or more software and/or hardware implementations of the present description.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims

1. A method of model training, comprising:

determining at least one additional feature from the base features;

2. The method of claim 1, wherein the user attribute data comprises: at least one of credit data of the user, insurance data of the user, medical data of the user.

3. The method of claim 1, wherein the historical first traffic data comprises: the business address comprises historical order data of the hospital address;

4. The method of claim 3, wherein determining additional features based on the number of order quantity intervals and feature data corresponding to the base features determined from the user attribute data comprises:

determining additional features from the plurality of clusters.

5. The method of claim 1, wherein the step of using the basic features and the at least one additional feature as wind control features and determining, for each wind control feature, a wind control data interval corresponding to the wind control feature comprises:

for each wind control feature, determining the relevance of the wind control feature on different data intervals, and for each data interval corresponding to the wind control feature, determining an information value IV corresponding to the wind control feature in the data interval according to the number of users of regular users and the number of users of reverse users, of which the feature data corresponding to the wind control feature is located in the data interval;

6. The method of claim 1, wherein the wind control model comprises: the system comprises a first wind control model and a second wind control model, wherein the first wind control model is used for carrying out service wind control in a first service stage of a second service, and the second wind control model is used for carrying out service wind control in a second service stage of the second service.

7. The method of claim 6, wherein the obtaining historical traffic data and user attribute data for each user if the first wind control model is trained comprises:

8. The method of claim 6, wherein the obtaining historical business data and user attribute data for each user if the second wind control model is trained comprises:

9. The method of claim 8, wherein before inputting, for each user, feature data located in the wind control data interval corresponding to each wind control feature, which is determined from the historical first service data and user attribute data of the user, into the wind control model to be trained and obtaining an output result, the method further comprises:

10. A method according to any of claims 7 to 9, wherein the first service phase of the second service comprises: and a second service stage of the second service comprises: and a claim settlement auditing stage.

11. A method for traffic scheduling, comprising:

receiving a service request of a user;

inputting the characteristic data into a pre-trained wind control model to obtain a wind control result, wherein the wind control model is obtained by training through the model training method of the claims 1-10;

12. An apparatus for model training, comprising:

13. An apparatus for traffic throttling, comprising:

the receiving module is used for receiving a service request of a user;

a wind control result determining module, configured to input the feature data into a pre-trained wind control model to obtain a wind control result, where the wind control model is obtained by training through the model training method according to any one of claims 1 to 10;

14. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-10 or 11.

15. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any of claims 1 to 10 or 11 when executing the program.