CN111160797A - Wind control model construction method and device, storage medium and terminal - Google Patents
Wind control model construction method and device, storage medium and terminal Download PDFInfo
- Publication number
- CN111160797A CN111160797A CN201911418593.2A CN201911418593A CN111160797A CN 111160797 A CN111160797 A CN 111160797A CN 201911418593 A CN201911418593 A CN 201911418593A CN 111160797 A CN111160797 A CN 111160797A
- Authority
- CN
- China
- Prior art keywords
- wind control
- control model
- characteristic information
- feature
- arrangement
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The embodiment of the invention discloses a method and a device for constructing a wind control model, a storage medium and a terminal. The method comprises the following steps: acquiring historical data of at least one user in a preset time period; extracting at least one type of first characteristic information related to historical data; sorting the first feature information in the historical data according to a preset sorting rule aiming at various types of first feature information to generate a first feature information sequence; determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence; and training a preset machine learning model based on the first arrangement quantile to generate a wind control model. By adopting the technical scheme, the arrangement quantiles of the characteristic information in the historical data replace the corresponding characteristic information, and model training is carried out on the basis of the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve the effect of dynamic self-adaption on the change of input data, and the speed of the wind control model decaying along with time is effectively reduced.
Description
Technical Field
The embodiment of the invention relates to the technical field of data processing, in particular to a method and a device for constructing a wind control model, a storage medium and a terminal.
Background
In the field of wind control, target user population and user characteristics of a wind control model are always changed due to the influences of macroscopic economic environment, company strategies and the like, so that the effect of the well-constructed wind control model is gradually reduced to be ineffective along with time. In the face of this situation, it is common to use new data to fine-tune the original model to meet the business requirements at intervals.
However, the scheme of adjusting the original wind control model by new data has the following two problems: 1. the modeling of new data in the near term can cause the new model to lose the information of the original model, the distribution of output scores generally has larger difference with the original model, and related personnel are required to perform detailed test and adjustment on the use strategy during use; 2. because the supervised model in the wind control field generally uses 3 months or 6 months of overdue days as a modeling target, the latest modeling data needs 3 months or 6 months of time to date, and the variability of the wind control data can be greatly changed within the time (3 months or 6 months), so that the adjustment of the wind control model has little effect.
Disclosure of Invention
The embodiment of the invention provides a method and a device for constructing a wind control model, a storage medium and a terminal, which are used for improving the timeliness of the wind control model.
In a first aspect, an embodiment of the present invention provides a method for constructing a wind control model, where the method includes:
acquiring historical data of at least one user in a preset time period;
extracting at least one type of first characteristic information related to the historical data;
sorting the first feature information in the historical data according to a preset sorting rule aiming at various types of first feature information to generate a first feature information sequence;
determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence;
and training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
In a second aspect, an embodiment of the present invention further provides a device for constructing a wind control model, where the device includes:
the historical data acquisition module is used for acquiring historical data of at least one user in a preset time period;
the first characteristic information extraction module is used for extracting at least one type of first characteristic information related to the historical data;
the first characteristic information sorting module is used for sorting the first characteristic information in the historical data according to a preset sorting rule aiming at various types of first characteristic information to generate a first characteristic information sequence;
the first arrangement quantile determining module is used for determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence;
and the wind control model generation module is used for training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a method for building a wind control model according to an embodiment of the present invention.
In a fourth aspect, an embodiment of the present invention provides a terminal, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor, when executing the computer program, implements the method for constructing a wind control model according to an embodiment of the present invention.
The method for constructing the wind control model, provided by the embodiment of the invention, comprises the steps of obtaining historical data of at least one user in a preset time period, extracting at least one type of first feature information related to the historical data, sequencing the first feature information in the historical data according to a preset sequencing rule aiming at the various types of first feature information to generate a first feature information sequence, determining a first sequencing quantile of the first feature information in the first feature information sequence, and finally training a preset machine learning model based on the first sequencing quantile to generate the wind control model. By adopting the technical means, the existing construction scheme of the wind control model is optimized, the corresponding characteristic information is replaced by the arrangement quantiles of the characteristic information in the historical data, and the model training is carried out based on the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve the effect of dynamic self-adaption on the change of input data, the time-dependent attenuation speed of the wind control model is effectively reduced, and the timeliness of the wind control model is further ensured.
Drawings
FIG. 1 is a schematic diagram of a construction process of a wind control model in the prior art;
fig. 2 is a schematic flow chart of a method for constructing a wind control model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating an update process of a wind control model in the prior art;
fig. 4 is a schematic flow chart of another method for constructing a wind control model according to an embodiment of the present invention;
fig. 5 is a schematic flow chart of another method for constructing a wind control model according to an embodiment of the present invention;
fig. 6 is a structural block diagram of a device for constructing a wind control model according to an embodiment of the present invention;
fig. 7 is a block diagram of a terminal according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Before discussing exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart may describe the steps as a sequential process, many of the steps can be performed in parallel, concurrently or simultaneously. In addition, the order of the steps may be rearranged. The process may be terminated when its operations are completed, but may have additional steps not included in the figure. The processes may correspond to methods, functions, procedures, subroutines, and the like.
In the prior art, features screened from original data are mainly and directly learned, so that a wind control model is generated. Fig. 1 is a schematic view of a construction process of a wind control model in the prior art. As shown in fig. 1, a series of processes such as data cleaning are performed on historical data, then characteristic information is screened from the processed historical data, the characteristic information is directly learned to generate a wind control model, and finally the wind control model is verified to be on-line, so that the wind control model is put into use; the data cleaning may include filtering out some data whose assignment is empty, and may also include desensitizing the historical data, that is, filtering out privacy information such as names, identification numbers, contact ways, and the like of the users in the historical data. Obviously, in the prior art, the wind control model is constructed by directly learning and generating the characteristic information on the dimension of the historical data. When risk prediction needs to be carried out on certain data or some data, the characteristics of the data are directly analyzed based on the constructed wind control model, and therefore a score value of the risk prediction is given. However, due to the variability of the wind control data, the input data is often changed greatly in data distribution, and the wind control model constructed based on the prior art cannot reflect the data change in time, so that the accuracy of the wind control model for predicting the risk of the data is lower and lower, that is, the timeliness of the wind control model is worse and worse. Therefore, the speed of the wind control model decaying along with time is reduced, and the timeliness of the wind control model is improved.
Fig. 2 is a flowchart of a method for building a wind control model according to an embodiment of the present invention, where the method may be performed by a device for building a wind control model, where the device may be implemented by software and/or hardware, and may be generally integrated in a terminal device. As shown in fig. 2, the method includes:
In the embodiment of the present invention, historical data of at least one user in a preset time period is obtained, where the historical data may include service data that has been determined to be legitimate or illegitimate (illegal). The length of the preset time period is not limited in the embodiment of the invention, and can be one month, two months, three months or half a year. For example, in an anti-money laundering scenario, the historical data may include account records of the user; as another example, there are often some false attention brushing or human behavior on the live platform, and in this scenario, the historical data may include related records of the user logging in the live platform, such as related data of registration, logging in, live watching, and bullet shooting. It should be noted that the embodiment of the present invention does not limit the type of the history data.
Optionally, extracting at least one feature information related to the historical data includes: acquiring at least one preset characteristic type; for each feature type, first feature information corresponding to the feature type is extracted from the history data. Illustratively, at least one preset feature type is obtained, wherein the feature type can be determined according to the data type of the historical data. For example, if the historical data is an account record, the preset feature types may include a charge-out amount, a charge-in amount, the number of account operations, and the like; for another example, the historical data is related data of a user logging in a live broadcast platform, and the preset feature types may include registration time, login times, live broadcast watching time, bullet screen launching times, and the like. For each feature type, first feature information corresponding to the feature type is extracted from the historical data, that is, the feature information corresponding to the feature type is determined from the historical data. Illustratively, the first feature information extracted from some historical data of the user a includes: the account number is 3 thousands, the account number is 5 thousands, and the account operation times are 3 times.
And 203, sorting the first characteristic information in the historical data according to a preset sorting rule aiming at various types of first characteristic information, and generating a first characteristic information sequence.
In the embodiment of the present invention, for each feature type, that is, for the feature type related to the first feature information, the first feature information corresponding to the same feature type in the historical data is sorted according to a preset sorting rule, so as to generate a first feature information sequence. For example, still taking the historical data as the account record as an example, the first feature information extracted from the historical data includes specific expenditure amount, expenditure amount and account operation times, the expenditure amounts in all the historical data are sorted according to a preset sorting rule to generate an expenditure amount sequence, the expenditure amounts in all the historical data are sorted to generate an expenditure amount sequence, and the account operation times in all the historical data are sorted to generate an account operation time sequence. For example, various types of first feature information may be sorted in order from small to large to generate a first feature information sequence; of course, the various types of first feature information may also be sorted in descending order to generate the first feature information sequence.
And 204, determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence.
In the embodiment of the present invention, a first arrangement quantile of the first feature information in the first feature sequence is determined, where the first arrangement quantile of the first feature information in the first feature sequence is a ratio of an arrangement position (a few bits) of the first feature information in the first feature sequence to the number of first feature information included in the first feature sequence. Illustratively, the first characteristic sequence generated based on the step 103 is a sequence generated by sorting the account operation times in a descending order, for example, the first characteristic sequence is [1,3,5,8,10], then the first ranking quantile of the account operation time 1 in the first characteristic sequence is 1/5, the first ranking quantile of the account operation time 3 in the first characteristic sequence is 2/5, the first ranking quantile of the account operation time 5 in the first characteristic sequence is 3/5, the first ranking quantile of the account operation time 8 in the first characteristic sequence is 4/5, and the first ranking quantile of the account operation time 10 in the first characteristic sequence is 1.
In the embodiment of the invention, a first arrangement quantile of first feature information corresponding to each feature type in a first feature information sequence can be used as a training sample to train a preset machine learning model to generate a wind control model. Optionally, the first ranking quantile determined in step 104 may replace the corresponding first feature information in the first feature information sequence to generate a first ranking quantile sequence, for example, the first ranking quantile sequence corresponding to the first feature sequence [1,3,5,8,10] is [1/5,2/5,3/5,4/5,1 ]. And then training a preset machine learning model based on the first arrangement quantile sequence to generate a wind control model. The preset machine learning model may be a convolutional neural network model, but it should be noted that the embodiment of the present invention does not limit the type of the preset machine learning model, and any algorithm that can train a model using a sequence as an input may be used as the preset machine learning model for training the wind control model.
In the embodiment of the invention, the characteristic information related to the characteristic types in the historical data is sequenced, the arrangement quantiles of the characteristic information in the characteristic information sequence replace the corresponding characteristic information, the arrangement quantiles are trained to generate the wind control model, the dimension of the input wind control model is changed, the wind control model focuses on the sequencing condition of each characteristic information in the historical data, and overfitting of the historical data can be avoided to a certain extent.
Optionally, the method further includes: and determining the corresponding relation between the first characteristic information and the first arrangement quantile in the first characteristic information sequence, and generating a first corresponding relation list. Optionally, after the generating the wind control model, the method further includes: acquiring prediction data to be subjected to wind control; searching a target arrangement quantile corresponding to the characteristic information related to the prediction data in the first corresponding relation list; and inputting the target arrangement quantiles into the wind control model, and determining whether the predicted data has risks according to the output result of the wind control model.
Illustratively, the first correspondence list is generated according to correspondence between the first arrangement quantile and the first feature information in the first feature information sequence. For example, a first correspondence list generated from the correspondence between the first arrangement quantile (1/5,2/5,3/5,4/5,1) and the first feature information of which the first feature sequence is [1,3,5,8,10] is as follows:
| first characteristic information | 1 | 3 | 5 | 8 | 10 |
| First permutation quantile | 1/5 | 2/5 | 3/5 | 4/5 | 1 |
In the embodiment of the invention, when risk control needs to be carried out on certain data or some data, the prediction data to be subjected to wind control is obtained, the characteristic information related to the prediction data is extracted, then the arrangement quantile corresponding to the characteristic information related to the prediction data is searched in the first corresponding relation list, and the searched arrangement quantile is used as the target arrangement quantile. The feature type corresponding to the feature information related to the prediction data is the same as the feature type corresponding to the first feature information related to the history data. Illustratively, if the number of account operations extracted from the prediction data is 3, finding a target ranking quantile 2/5 in the first correspondence list. And finally, inputting the target arrangement quantiles into the trained wind control model, and determining whether the prediction data has risks according to the output result of the wind control model. The wind control model can output a score value, and when the score value is larger than a preset score threshold value, the prediction data can be determined to have risks; when the score value is less than a preset score threshold, it may be determined that the predictive data is not at risk.
The method for constructing the wind control model, provided by the embodiment of the invention, comprises the steps of obtaining historical data of at least one user in a preset time period, extracting at least one type of first feature information related to the historical data, sequencing the first feature information in the historical data according to a preset sequencing rule aiming at the various types of first feature information to generate a first feature information sequence, determining a first sequencing quantile of the first feature information in the first feature information sequence, and finally training a preset machine learning model based on the first sequencing quantile to generate the wind control model. By adopting the technical means, the existing construction scheme of the wind control model is optimized, the corresponding characteristic information is replaced by the arrangement quantiles of the characteristic information in the historical data, and the model training is carried out based on the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve the effect of dynamic self-adaption on the change of input data, the time-dependent attenuation speed of the wind control model is effectively reduced, and the timeliness of the wind control model is further ensured.
In order to ensure the accuracy of the risk prediction of the wind control model, the wind control model needs to be updated regularly or irregularly. Fig. 3 is a schematic diagram of an update process of a wind control model in the prior art. As shown in fig. 3, in the prior art, when a wind control model is updated, an original trained wind control model is trained to perform modeling anew based on update data with labels (legal or illegal), which can make the updated wind control model more truly reflect the risk condition of the update data, but the original data for building the wind control model is not preserved, and when the risk model is updated in the prior art, label information of the update data needs to be obtained, which generally needs to wait for a period of time (e.g., three months), so that the update data for updating the wind control model has a period of time from the current time, so that the wind control model cannot be built by using the latest user data as a sample, and a certain delay always exists.
In some embodiments, after generating the wind control model, the method further includes: when detecting that a wind control model updating event is triggered, acquiring updating data of a wind control model; extracting at least one type of second characteristic information related to the updating data; the feature type corresponding to the second feature information is the same as the feature type corresponding to the first feature information, and the number of the feature types corresponding to the second feature information is the same as the number of the feature types corresponding to the first feature information; aiming at various types of second feature information, sorting the second feature information in the updated data according to the preset sorting rule to generate a second feature information sequence; determining a second arrangement quantile of the second characteristic information in the second characteristic information sequence; determining the corresponding relation between the second characteristic information and the second arrangement quantile in the second characteristic information sequence, and generating a second corresponding relation list; updating the first correspondence list based on the second correspondence list.
For example, in order to update the wind control model at an appropriate timing, a condition that the wind control model update event is triggered may be set in advance. Optionally, in order to really determine the updating requirement of the user on the wind control model, when it is detected that the current user actively opens the updating right of the wind control model, an updating event of the wind control model may be triggered. Optionally, in order to enable the wind control model to predict the risk more accurately, an update event of the wind control model may be triggered every preset time, for example, the wind control model is updated every six months. It should be noted that, the embodiment of the present application does not limit the specific representation form in which the update event of the wind control model is triggered.
And when detecting that the update event of the wind control model is triggered, acquiring the update data of the wind control model. The type of the updated data of the wind control model is the same as the data type of the historical data, for example, the updated data is a bill record, and the updated data is related data of a user logging in a live broadcast platform. Extracting at least one type of second feature information related to the update data, optionally, extracting at least one type of second feature information related to the update data, includes: and acquiring at least one preset feature type, and extracting second feature information corresponding to the feature type from the updated data for each feature type. Then, according to the feature type related to the second feature information, sorting the second feature information corresponding to the same feature type in the updated data according to a preset sorting rule, and generating a second feature information sequence. The arrangement mode of the second characteristic information is the same as that of the first characteristic information. And determining a second arrangement quantile of the second characteristic information in the second characteristic sequence, wherein the second arrangement quantile of the second characteristic information in the second characteristic sequence is the ratio of the arrangement position (the number of bits) of the second characteristic information in the second characteristic sequence to the number of the second characteristic information contained in the second characteristic sequence. And generating a second corresponding relation list according to the corresponding relation between the second arrangement quantile and the second characteristic information in the second characteristic sequence.
For example, the second feature sequence is also a sequence generated by sorting the number of operation on the account in the order from small to large, for example, the second feature sequence is [3,6,7,10,12], and a second correspondence list generated according to a correspondence between the second arrangement quantile (1/5,2/5,3/5,4/5,1) and the second feature information in the second feature sequence [3,6,7,10,12] is as follows:
| second characteristic information | 3 | 6 | 7 | 10 | 12 |
| Second permutation quantile | 1/5 | 2/5 | 3/5 | 4/5 | 1 |
And updating the first correspondence list with the second correspondence list. It can be understood that when risk prediction is performed on certain data or some data based on the wind control model, the target arrangement quantiles corresponding to the feature information related to the predicted data are directly searched in the second corresponding relation list, the target arrangement quantiles are input into the wind control model, and whether risk exists in the predicted data is determined according to the output result of the wind control model. For example, if the number of account operations extracted from the prediction data is 3, finding a target ranking score of 1/5 in the second correspondence list (that is, the second correspondence list updated based on the second correspondence list), and inputting 1/5 to the wind control model for risk prediction.
For another example, the second signature sequence is also generated by sorting the number of operation on the account in descending order, for example, the second signature sequence is [1,3,5,6,8,9,10,13], and the second correspondence relationship list generated according to the correspondence relationship between the second arrangement quantile (1/8,2/8,3/8,4/8,5/8,6/8,7/8,1) and the second signature information in the second signature sequence [1,3,5,6,8,9,10,13] is as follows:
| second characteristic information | 1 | 3 | 5 | 6 | 8 | 9 | 10 | 13 |
| Second permutation quantile | 1/8 | 2/8 | 3/8 | 4/8 | 5/8 | 6/8 | 7/8 | 1 |
For example, if the number of account operations extracted from the prediction data is 3, finding a target ranking score of 2/8 in the second correspondence list (that is, the second correspondence list updated based on the second correspondence list), and inputting 2/8 to the wind control model for risk prediction.
It should be noted that, in the embodiment of the present invention, the data volumes of the update data and the historical data are not limited, and the data volume of the update data and the data volume of the historical data may be the same or different.
It can be understood that when the wind control model needs to be updated, the originally constructed wind control model does not need to be trained based on the update data to perform modeling again, the first corresponding relationship list constructed by the historical data can be directly updated by the second corresponding relationship list constructed by the update data, but the originally constructed wind control model is kept unchanged, so that the wind control model is indirectly updated. It can be understood that when risk prediction needs to be performed on prediction data, a target arrangement quantile is directly searched based on the updated first corresponding relationship list (i.e., the second corresponding relationship list), and then the target arrangement quantile is input into the originally constructed wind control model. The advantage of this arrangement is that, on the one hand, not only the relevant information of the historical data can be retained, but also an effective combination of the updated data (new data) and the historical data (original data) can be realized; on the other hand, when the wind control model is updated, the wind control model can be rapidly updated without waiting for a certain time such as 3 months or 6 months and determining the label (legal data or illegal data) of the updated data, so that the wind control model has better adaptability when risk prediction is performed.
Fig. 4 is a schematic flow chart of another method for constructing a wind control model according to an embodiment of the present invention, and as shown in fig. 4, the method includes the following steps:
And step 404, sorting the first characteristic information in the historical data according to a preset sorting rule aiming at various types of first characteristic information, and generating a first characteristic information sequence.
And 406, training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
And step 408, when the wind control event is detected to be triggered, acquiring prediction data to be subjected to wind control.
And step 409, searching a target arrangement quantile corresponding to the characteristic information related to the prediction data in the first corresponding relation list.
And step 410, inputting the target arrangement quantiles into a wind control model, and determining whether the predicted data has risks according to the output result of the wind control model.
According to the method for constructing the wind control model, the arrangement quantiles of the characteristic information in the historical data are used for replacing the corresponding characteristic information, model training is carried out on the basis of the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve a dynamic self-adaptive effect on the change of input data, the time-dependent attenuation speed of the wind control model is effectively reduced, and the timeliness of the wind control model is further guaranteed.
Fig. 5 is a schematic flow chart of another method for constructing a wind control model according to an embodiment of the present invention, and as shown in fig. 5, the method includes the following steps:
And 505, determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence.
And 506, training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
And 507, determining the corresponding relation between the first characteristic information in the first characteristic information sequence and the first arrangement quantile, and generating a first corresponding relation list.
The feature types corresponding to the second feature information are the same as the feature types corresponding to the first feature information, and the number of the feature types corresponding to the second feature information is the same as the number of the feature types corresponding to the first feature information.
And 510, aiming at various types of second characteristic information, sequencing the second characteristic information in the updated data according to a preset sequencing rule to generate a second characteristic information sequence.
And 511, determining a second arrangement quantile of the second characteristic information in the second characteristic information sequence.
And step 512, determining the corresponding relation between the second characteristic information in the second characteristic information sequence and the second arrangement quantile, and generating a second corresponding relation list.
Optionally, after step 513, when risk prediction needs to be performed on some data based on the wind control model, a target arrangement quantile corresponding to feature information related to the predicted data is directly searched in the second correspondence list, the target arrangement quantile is input into the wind control model, and whether risk exists in the predicted data is determined according to an output result of the wind control model.
According to the method for constructing the wind control model, the arrangement quantiles of the characteristic information in the historical data are used for replacing the corresponding characteristic information, model training is carried out on the basis of the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve a dynamic self-adaptive effect on the change of input data, the time-dependent attenuation speed of the wind control model is effectively reduced, and the timeliness of the wind control model is further guaranteed. When the wind control model needs to be updated, the originally constructed wind control model does not need to be trained based on the updating data to be modeled again, the first corresponding relation list constructed by the historical data can be directly updated by the second corresponding relation list constructed by the updating data, but the originally constructed wind control model is kept unchanged, so that the updating of the wind control model is indirectly realized, on one hand, the relevant information of the historical data can be kept, and the effective combination of the updating data and the historical data can be realized; on the other hand, when the wind control model is updated, the wind control model can be rapidly updated without waiting for a certain time such as 3 months or 6 months and determining the label (legal data or illegal data) of the updated data, so that the wind control model has better adaptability when risk prediction is performed.
Fig. 6 is a structural block diagram of a device for building a wind control model according to an embodiment of the present invention, where the device may be implemented by software and/or hardware, and is generally integrated in a terminal, and the device may be configured to build the wind control model by executing a method for building the wind control model. As shown in fig. 6, the apparatus includes:
a historical data obtaining module 601, configured to obtain historical data of at least one user in a preset time period;
a first feature information extraction module 602, configured to extract at least one type of first feature information related to the historical data;
the first feature information sorting module 603 is configured to sort, according to a preset sorting rule, the first feature information in the historical data according to various types of first feature information, and generate a first feature information sequence;
a first ranking quantile determining module 604, configured to determine a first ranking quantile of the first feature information in the first feature information sequence;
and the wind control model generating module 605 is configured to train a preset machine learning model based on the first arrangement quantile to generate a wind control model.
The construction device of the wind control model provided by the embodiment of the invention is used for acquiring historical data of at least one user in a preset time period, extracting at least one type of first characteristic information related to the historical data, sequencing the first characteristic information in the historical data according to a preset sequencing rule aiming at the various types of first characteristic information to generate a first characteristic information sequence, determining a first sequencing quantile of the first characteristic information in the first characteristic information sequence, and finally training a preset machine learning model based on the first sequencing quantile to generate the wind control model. By adopting the technical means, the existing construction scheme of the wind control model is optimized, the corresponding characteristic information is replaced by the arrangement quantiles of the characteristic information in the historical data, and the model training is carried out based on the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve the effect of dynamic self-adaption on the change of input data, the time-dependent attenuation speed of the wind control model is effectively reduced, and the timeliness of the wind control model is further ensured.
Optionally, the first feature information extraction module is configured to:
acquiring at least one preset characteristic type;
for each feature type, first feature information corresponding to the feature type is extracted from the history data.
Optionally, the apparatus further comprises:
and the first corresponding relation list generating module is used for determining the corresponding relation between the first characteristic information in the first characteristic information sequence and the first arrangement quantile and generating a first corresponding relation list.
Optionally, the apparatus further comprises:
the prediction data acquisition module is used for acquiring prediction data to be subjected to wind control after the wind control model is generated;
a ranking quantile search module, configured to search, in the first correspondence list, a target ranking quantile corresponding to the feature information related to the prediction data;
and the wind control model prediction module is used for inputting the target arrangement quantiles into the wind control model and determining whether the predicted data has risks according to the output result of the wind control model.
Optionally, the apparatus further comprises:
the updating data acquisition module is used for acquiring updating data of the wind control model when detecting that a wind control model updating event is triggered after the wind control model is generated;
the second characteristic information extraction module is used for extracting at least one type of second characteristic information related to the updating data; the feature type corresponding to the second feature information is the same as the feature type corresponding to the first feature information, and the number of the feature types corresponding to the second feature information is the same as the number of the feature types corresponding to the first feature information;
the second characteristic information sequence generation module is used for sequencing the second characteristic information in the updated data according to the preset sequencing rule aiming at various types of second characteristic information to generate a second characteristic information sequence;
a second arrangement quantile determining module, configured to determine a second arrangement quantile of the second feature information in the second feature information sequence;
a second correspondence list generation module, configured to determine a correspondence between the second feature information in the second feature information sequence and the second arrangement quantile, and generate a second correspondence list;
and the corresponding relation list updating module is used for updating the first corresponding relation list based on the second corresponding relation list.
Embodiments of the present invention also provide a storage medium containing computer-executable instructions, which when executed by a computer processor, are configured to perform a method for building a wind control model, where the method includes:
acquiring historical data of at least one user in a preset time period;
extracting at least one type of first characteristic information related to the historical data;
sorting the first feature information in the historical data according to a preset sorting rule aiming at various types of first feature information to generate a first feature information sequence;
determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence;
and training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
Storage medium-any of various types of memory devices or storage devices. The term "storage medium" is intended to include: mounting media such as CD-ROM, floppy disk, or tape devices; computer system memory or random access memory such as DRAM, DDRRAM, SRAM, EDORAM, Lanbas (Rambus) RAM, etc.; non-volatile memory such as flash memory, magnetic media (e.g., hard disk or optical storage); registers or other similar types of memory elements, etc. The storage medium may also include other types of memory or combinations thereof. In addition, the storage medium may be located in a first computer system in which the program is executed, or may be located in a different second computer system connected to the first computer system through a network (such as the internet). The second computer system may provide program instructions to the first computer for execution. The term "storage medium" may include two or more storage media that may reside in different locations, such as in different computer systems that are connected by a network. The storage medium may store program instructions (e.g., embodied as a computer program) that are executable by one or more processors.
Of course, the storage medium containing the computer-executable instructions provided by the embodiments of the present invention is not limited to the above-described building operation of the wind control model, and may also perform related operations in the building method of the wind control model provided by any embodiment of the present invention.
The embodiment of the invention provides a terminal, and a device for constructing a wind control model provided by the embodiment of the invention can be integrated in the terminal. Fig. 7 is a block diagram of a terminal according to an embodiment of the present invention. The terminal 700 may include: the wind control model building method comprises a memory 701, a processor 702 and a computer program which is stored on the memory 701 and can be run by the processor, wherein the processor 702 realizes the building method of the wind control model according to the embodiment of the invention when executing the computer program.
The terminal provided by the embodiment of the invention obtains the historical data of at least one user in a preset time period, extracts at least one type of first feature information related to the historical data, sorts the first feature information in the historical data according to a preset sorting rule aiming at the various types of first feature information to generate a first feature information sequence, determines a first sorting quantile of the first feature information in the first feature information sequence, and finally trains a preset machine learning model based on the first sorting quantile to generate the wind control model. By adopting the technical means, the existing construction scheme of the wind control model is optimized, the corresponding characteristic information is replaced by the arrangement quantiles of the characteristic information in the historical data, and the model training is carried out based on the arrangement quantiles of the characteristic information to generate the wind control model, so that the constructed wind control model can achieve the effect of dynamic self-adaption on the change of input data, the time-dependent attenuation speed of the wind control model is effectively reduced, and the timeliness of the wind control model is further ensured.
The device, the storage medium and the terminal for constructing the wind control model provided in the above embodiments can execute the method for constructing the wind control model provided in any embodiment of the present invention, and have corresponding functional modules and beneficial effects for executing the method. For technical details that are not described in detail in the above embodiments, reference may be made to a method for constructing a wind control model according to any embodiment of the present invention.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (10)
1. A method for constructing a wind control model is characterized by comprising the following steps:
acquiring historical data of at least one user in a preset time period;
extracting at least one type of first characteristic information related to the historical data;
sorting the first feature information in the historical data according to a preset sorting rule aiming at various types of first feature information to generate a first feature information sequence;
determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence;
and training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
2. The method of claim 1, wherein extracting at least one feature information related to the historical data comprises:
acquiring at least one preset characteristic type;
for each feature type, first feature information corresponding to the feature type is extracted from the history data.
3. The method of claim 1, further comprising:
and determining the corresponding relation between the first characteristic information and the first arrangement quantile in the first characteristic information sequence, and generating a first corresponding relation list.
4. The method of claim 3, after generating the wind control model, further comprising:
acquiring prediction data to be subjected to wind control;
searching a target arrangement quantile corresponding to the characteristic information related to the prediction data in the first corresponding relation list;
and inputting the target arrangement quantiles into the wind control model, and determining whether the predicted data has risks according to the output result of the wind control model.
5. The method of claim 3, after generating the wind control model, further comprising:
when detecting that a wind control model updating event is triggered, acquiring updating data of a wind control model;
extracting at least one type of second characteristic information related to the updating data; the feature type corresponding to the second feature information is the same as the feature type corresponding to the first feature information, and the number of the feature types corresponding to the second feature information is the same as the number of the feature types corresponding to the first feature information;
aiming at various types of second feature information, sorting the second feature information in the updated data according to the preset sorting rule to generate a second feature information sequence;
determining a second arrangement quantile of the second characteristic information in the second characteristic information sequence;
determining the corresponding relation between the second characteristic information and the second arrangement quantile in the second characteristic information sequence, and generating a second corresponding relation list;
updating the first correspondence list based on the second correspondence list.
6. A wind control model building device is characterized by comprising:
the historical data acquisition module is used for acquiring historical data of at least one user in a preset time period;
the first characteristic information extraction module is used for extracting at least one type of first characteristic information related to the historical data;
the first characteristic information sorting module is used for sorting the first characteristic information in the historical data according to a preset sorting rule aiming at various types of first characteristic information to generate a first characteristic information sequence;
the first arrangement quantile determining module is used for determining a first arrangement quantile of the first characteristic information in the first characteristic information sequence;
and the wind control model generation module is used for training a preset machine learning model based on the first arrangement quantile to generate a wind control model.
7. The apparatus of claim 6, further comprising:
and the first corresponding relation list generating module is used for determining the corresponding relation between the first characteristic information in the first characteristic information sequence and the first arrangement quantile and generating a first corresponding relation list.
8. The apparatus of claim 7, further comprising:
the updating data acquisition module is used for acquiring updating data of the wind control model when detecting that a wind control model updating event is triggered after the wind control model is generated;
the second characteristic information extraction module is used for extracting at least one type of second characteristic information related to the updating data; the feature type corresponding to the second feature information is the same as the feature type corresponding to the first feature information, and the number of the feature types corresponding to the second feature information is the same as the number of the feature types corresponding to the first feature information;
the second characteristic information sequence generation module is used for sequencing the second characteristic information in the updated data according to the preset sequencing rule aiming at various types of second characteristic information to generate a second characteristic information sequence;
a second arrangement quantile determining module, configured to determine a second arrangement quantile of the second feature information in the second feature information sequence;
a second correspondence list generation module, configured to determine a correspondence between the second feature information in the second feature information sequence and the second arrangement quantile, and generate a second correspondence list;
and the corresponding relation list updating module is used for updating the first corresponding relation list based on the second corresponding relation list.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out a method of constructing a wind control model according to any one of claims 1 to 5.
10. A terminal comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method of constructing a wind control model according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911418593.2A CN111160797A (en) | 2019-12-31 | 2019-12-31 | Wind control model construction method and device, storage medium and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911418593.2A CN111160797A (en) | 2019-12-31 | 2019-12-31 | Wind control model construction method and device, storage medium and terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111160797A true CN111160797A (en) | 2020-05-15 |
Family
ID=70560371
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911418593.2A Pending CN111160797A (en) | 2019-12-31 | 2019-12-31 | Wind control model construction method and device, storage medium and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111160797A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111861701A (en) * | 2020-07-09 | 2020-10-30 | 深圳市富之富信息技术有限公司 | Wind control model optimization method and device, computer equipment and storage medium |
| CN112966131A (en) * | 2021-03-02 | 2021-06-15 | 中华人民共和国成都海关 | Customs data wind control type identification method, customs intelligent risk distribution and control device, computer equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305158A (en) * | 2017-12-27 | 2018-07-20 | 阿里巴巴集团控股有限公司 | A kind of method, apparatus and equipment of trained air control model and air control |
| CN108399477A (en) * | 2017-02-07 | 2018-08-14 | 阿里巴巴集团控股有限公司 | A kind of method and device that risk threshold value determines |
| CN108549954A (en) * | 2018-03-26 | 2018-09-18 | 平安科技（深圳）有限公司 | Risk model training method, risk identification method, device, equipment and medium |
-
2019
- 2019-12-31 CN CN201911418593.2A patent/CN111160797A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108399477A (en) * | 2017-02-07 | 2018-08-14 | 阿里巴巴集团控股有限公司 | A kind of method and device that risk threshold value determines |
| CN108305158A (en) * | 2017-12-27 | 2018-07-20 | 阿里巴巴集团控股有限公司 | A kind of method, apparatus and equipment of trained air control model and air control |
| CN108549954A (en) * | 2018-03-26 | 2018-09-18 | 平安科技（深圳）有限公司 | Risk model training method, risk identification method, device, equipment and medium |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111861701A (en) * | 2020-07-09 | 2020-10-30 | 深圳市富之富信息技术有限公司 | Wind control model optimization method and device, computer equipment and storage medium |
| CN112966131A (en) * | 2021-03-02 | 2021-06-15 | 中华人民共和国成都海关 | Customs data wind control type identification method, customs intelligent risk distribution and control device, computer equipment and storage medium |
| CN112966131B (en) * | 2021-03-02 | 2022-09-16 | 中华人民共和国成都海关 | Customs data wind control type identification method, customs intelligent risk distribution and control device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109922032B (en) | Method, device, equipment and storage medium for determining risk of logging in account | |
| CN111369299B (en) | Identification method, device, equipment and computer readable storage medium | |
| CN107071193B (en) | Method and device for accessing interactive response system to user | |
| CN111428231A (en) | Safety processing method, device and equipment based on user behaviors | |
| CN106408411A (en) | Credit assessment method and device | |
| CN110570312B (en) | Sample data acquisition method and device, computer equipment and readable storage medium | |
| EP3726441A1 (en) | Company bankruptcy prediction system and operating method therefor | |
| CN110728301A (en) | Credit scoring method, device, terminal and storage medium for individual user | |
| CN111160797A (en) | Wind control model construction method and device, storage medium and terminal | |
| CN113449753B (en) | Service risk prediction method, device and system | |
| CN115329746A (en) | Event extraction method, device and equipment | |
| CN112035775B (en) | User identification method and device based on random forest model and computer equipment | |
| CN111325255B (en) | Specific crowd delineating method and device, electronic equipment and storage medium | |
| US11176502B2 (en) | Analytical model training method for customer experience estimation | |
| CN112488163A (en) | Abnormal account identification method and device, computer equipment and storage medium | |
| CN111582313A (en) | Sample data generation method and device and electronic equipment | |
| CN113596061B (en) | Network security vulnerability response method based on block chain technology | |
| CN114581086A (en) | Phishing account detection method and system based on dynamic time sequence network | |
| CN103942403A (en) | Method and device for screening mass variables | |
| CN113628077A (en) | Method for generating non-repeated examination questions, terminal and readable storage medium | |
| CN111104963A (en) | Target user determination method and device, storage medium and electronic equipment | |
| CN112069485B (en) | Safety processing method, device and equipment based on user behaviors | |
| CN116962089B (en) | Network monitoring method and system for information security | |
| CN114978616B (en) | Construction method and device of risk assessment system, and risk assessment method and device | |
| CN111369352B (en) | Joint modeling method, apparatus, and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |