CN113112347A - Determination method of hasty collection decision, related device and computer storage medium - Google Patents
Determination method of hasty collection decision, related device and computer storage medium Download PDFInfo
- Publication number
- CN113112347A CN113112347A CN202110494902.5A CN202110494902A CN113112347A CN 113112347 A CN113112347 A CN 113112347A CN 202110494902 A CN202110494902 A CN 202110494902A CN 113112347 A CN113112347 A CN 113112347A
- Authority
- CN
- China
- Prior art keywords
- data
- decision
- model
- information
- collection
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Abstract
The application provides a determination method of a hasty decision, a related device and a computer storage medium, wherein the determination method of the hasty decision comprises the following steps: firstly, acquiring debt information of a target client; then, performing preset index extraction on the debt information to obtain at least one target index information; finally, inputting all the target index information into a collection prompting decision determining model, and outputting to obtain a collection prompting decision for the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models. So as to achieve the purpose of quickly and accurately determining the hastening decision aiming at the target client.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method for determining a collection decision, a related apparatus, and a computer storage medium.
Background
At present, domestic macro economy is in a slow down channel, and commercial banks all increase the development of retail business. The business of personal loans, especially short-term consumption loans and credit cards, is accelerated and the personal loan balance continues to increase. However, the individual poor assets have the characteristic of small amount dispersion, and are greatly different from the traditional poor asset treatment.
In the prior art, for a large number of bad property projects of personal loans, business personnel mainly know the project conditions one by one, adopt different accepting and handling strategies according to experience, temporarily cannot obtain decision information quickly and efficiently through a system, and often need to consume a large amount of time and energy in the face of the characteristics of large number and type of bad personal loans.
Disclosure of Invention
In view of the foregoing, the present application provides a method, an apparatus and a computer storage medium for determining a prompt receipt decision, which are used to quickly and accurately determine a prompt receipt decision for a target client.
In a first aspect, the present application provides a method for determining a hasty decision, including:
acquiring debt information of a target client;
performing index extraction on the debt information to obtain at least one target index information;
inputting all the target index information into a collection prompting decision determining model, and outputting to obtain a collection prompting decision for the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models.
Optionally, the method for constructing the incentive decision determination model includes:
constructing a training sample set; wherein the training sample set comprises historical debt information of at least one training sample client and the sensitivity of the training sample client for each charging strategy;
inputting the historical debt information of the training sample client into a plurality of two-classification prediction models in a preset model to obtain a prediction result aiming at each collection strategy;
for each collection strategy, predicting the effectiveness of the collection strategy by using an intelligent decision model in the preset model to obtain a first prediction result; the intelligent decision model is a logistic regression model;
calibrating the first prediction result by using an expert rule model in the preset model to obtain a second prediction result; the expert rule model is a preset rule model;
and adjusting the preset model by utilizing the second prediction result of the collection prompting strategy and the error of the training sample client for the sensitivity of the collection prompting strategy until the error of the adjusted preset model output for the second prediction result of the collection prompting strategy and the error of the training sample client for the sensitivity of the collection prompting strategy meet the preset convergence condition, and taking the adjusted preset model as a collection prompting decision determination model.
Optionally, the constructing a training sample set includes:
acquiring historical debt information of a training sample client;
index extraction is carried out on the historical debt information of the training sample client to obtain at least one sample index information;
preprocessing each sample index information to obtain target preprocessing data;
respectively formulating a corresponding rule for each sample index information, and setting a corresponding collection and disposal strategy for each sample index group; wherein the set of sample metrics includes at least one of the sample metric information;
determining a sensitivity of a training sample customer to each of the revenue handling policies.
Optionally, the preprocessing each sample index information to obtain target preprocessing data includes:
for each sample index information, preprocessing structured data in the sample index information to obtain first preprocessed data;
for each piece of sample index information, preprocessing unstructured data in the sample index information to obtain second preprocessed data;
and taking all the first preprocessing data and all the second preprocessing data as target preprocessing data.
Optionally, for each piece of sample index information, preprocessing structured data in the sample index information to obtain first preprocessed data includes:
for each piece of the structured data, judging whether the missing value of the structured data is larger than a first threshold value;
if the missing value of the structured data is judged to be larger than a first threshold value, deleting the structured data;
if the missing value of the structured data is judged to be not larger than a first threshold value, judging whether the structured data is a continuous variable;
if the structured data are judged to be continuous variables, carrying out box separation on the structured data;
selecting information values of the boxed structured data to obtain first preprocessing data;
if the structured data is judged not to be a continuous variable, judging whether the level number of the structured data is larger than a second threshold value;
if the level number of the structured data is judged to be larger than a second threshold value, performing variable clustering on the structured data;
selecting information values of the structured data after variable clustering to obtain first preprocessing data;
and if the level number of the structured data is judged to be not larger than a second threshold value, selecting an information value of the structured data to obtain first preprocessing data.
Optionally, the method for determining a prompt-to-withdraw decision further includes:
optimizing all the bins;
removing the structured data with co-linearity in each of the bins.
Optionally, for each piece of sample index information, preprocessing unstructured data in the sample index information to obtain second preprocessed data includes:
performing corpus cleaning on the unstructured data to obtain first target data;
performing word segmentation processing on the first target data to obtain second target data;
extracting keywords from the second target data to obtain third target data;
and performing feature extraction on the third target data to obtain second preprocessing data.
A second aspect of the present application provides a determination device for an incentive decision, comprising:
the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring debt information of a target client;
the first extraction unit is used for carrying out index extraction on the debt information to obtain at least one target index information;
the first determining unit is used for inputting all the target index information into a collection decision determining model and outputting the collection decision aiming at the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models.
Optionally, the building unit of the incentive decision determining model includes:
the training sample set constructing unit is used for constructing a training sample set; wherein the training sample set comprises historical debt information of at least one training sample client and the sensitivity of the training sample client for each charging strategy;
the input unit is used for inputting the historical debt information of the training sample client into a plurality of two-classification prediction models in a preset model to obtain a prediction result aiming at each collection urging strategy;
the prediction unit is used for predicting the effectiveness of the collection strategies by utilizing an intelligent decision model in the preset model aiming at each collection strategy to obtain a first prediction result; the intelligent decision model is a logistic regression model;
the calibration unit is used for calibrating the first prediction result by utilizing an expert rule model in the preset model to obtain a second prediction result; the expert rule model is a preset rule model;
and the adjusting unit is used for adjusting the preset model by utilizing the second prediction result of the collection prompting strategy and the error of the training sample client for the sensitivity of the collection prompting strategy until the error of the adjusted preset model output for the second prediction result of the collection prompting strategy and the error of the training sample client for the sensitivity of the collection prompting strategy meet the preset convergence condition, and taking the adjusted preset model as a collection prompting decision determining model.
Optionally, the sample set constructing unit includes:
the second acquisition unit is used for acquiring historical debt information of the training sample client;
the second extraction unit is used for performing index extraction on the historical debt information of the training sample client to obtain at least one sample index information;
the preprocessing unit is used for preprocessing each sample index information to obtain target preprocessing data;
the setting unit is used for respectively making a corresponding rule for each sample index information and setting a corresponding collection and disposal strategy for each sample index group; wherein the set of sample metrics includes at least one of the sample metric information;
a second determining unit, configured to determine a sensitivity of a training sample client to each of the revenue handling policies.
Optionally, the preprocessing unit includes:
the first preprocessing subunit is used for preprocessing the structured data in the sample index information to obtain first preprocessed data aiming at each sample index information;
the second preprocessing subunit is used for preprocessing unstructured data in the sample index information aiming at each piece of sample index information to obtain second preprocessed data;
and a third determining unit, configured to use all the first preprocessed data and all the second preprocessed data as target preprocessed data.
Optionally, the first preprocessing subunit includes:
a first judging unit, configured to judge, for each of the structured data, whether a missing value of the structured data is greater than a first threshold;
a deleting unit, configured to delete the structured data if the first determining unit determines that the missing value of the structured data is greater than a first threshold;
a second judging unit, configured to judge whether the structured data is a continuous variable if the first judging unit judges that the missing value of the structured data is not greater than a first threshold;
a binning unit, configured to bin the structured data if the structured data is determined to be a continuous variable by the second determining unit;
the information value selection unit is used for selecting the information values of the boxed structured data to obtain first preprocessing data;
a third judging unit, configured to judge whether the level number of the structured data is greater than a second threshold value if the structured data is not a continuous variable as judged by the second judging unit;
the variable clustering unit is used for performing variable clustering on the structured data if the level number of the structured data is larger than a second threshold value, which is judged by the third judging unit;
the information value selection unit is also used for selecting the information value of the structured data after variable clustering to obtain first preprocessing data;
the information value selecting unit is further configured to, if the third determining unit determines that the level number of the structured data is not greater than a second threshold, perform information value selection on the structured data to obtain first preprocessed data.
Optionally, the determining device for a catalyst recovery decision further includes:
the optimization unit is used for optimizing all the sub-boxes;
and the removing unit is used for removing the structured data with the collinearity in each of the boxes.
Optionally, the second preprocessing subunit includes:
the cleaning unit is used for obtaining first target data after corpus cleaning is carried out on the unstructured data;
the word segmentation unit is used for carrying out word segmentation processing on the first target data to obtain second target data;
the keyword extraction unit is used for extracting keywords from the second target data to obtain third target data;
and the feature extraction unit is used for extracting features of the third target data to obtain second preprocessing data.
A third aspect of the present application provides a server comprising:
one or more processors;
a storage device having one or more programs stored thereon;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of determining an hasten decision as defined in any of the first aspects.
A fourth aspect of the present application provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for determining an hasten decision according to any one of the first aspect.
In view of the foregoing, in the present application, there is provided a method for determining an incentive decision, a related apparatus, and a computer storage medium, the method for determining an incentive decision, including: firstly, acquiring debt information of a target client; then, performing index extraction on the debt information to obtain at least one target index information; finally, inputting all the target index information into a collection prompting decision determining model, and outputting to obtain a collection prompting decision for the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models. So as to achieve the purpose of quickly and accurately determining the hastening decision aiming at the target client.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flowchart illustrating an exemplary method for determining a catalyst decision according to an embodiment of the present disclosure;
FIG. 2 is a flow chart of a method for constructing a pull-in decision-making model according to another embodiment of the present disclosure;
FIG. 3 is a flowchart of a method for constructing a training sample set according to another embodiment of the present application;
fig. 4 is a flowchart of a method for preprocessing sample index information according to another embodiment of the present disclosure;
FIG. 5 is a flow chart of a method for preprocessing structured data according to another embodiment of the present application;
FIG. 6 is a flow chart of a method for preprocessing unstructured data according to another embodiment of the present application;
FIG. 7 is a schematic diagram of an apparatus for determining a catalyst recovery decision according to another embodiment of the present disclosure;
fig. 8 is a schematic diagram of a server implementing a determination method for a hasty check-out decision according to another embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first", "second", and the like, referred to in this application, are only used for distinguishing different devices, modules or units, and are not used for limiting the order or interdependence of functions performed by these devices, modules or units, but the terms "include", or any other variation thereof are intended to cover a non-exclusive inclusion, so that a process, method, article, or apparatus that includes a series of elements includes not only those elements but also other elements that are not explicitly listed, or includes elements 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 identical elements in a process, method, article, or apparatus that comprises the element.
The embodiment of the application provides a method for determining a hasty decision, which specifically includes the following steps as shown in fig. 1:
s101, acquiring debt information of a target client.
The debt information includes, but is not limited to, a personal loan operation handling ledger, a historical collection record, a bad personal loan detail ledger, a personal loan outside commission collection information, a personal loan litigation information, a personal loan detail, a loan basic information, a personal loan repayment flow, a judicial complaint information, a credit investigation information, a distrusted customer list, an operator data, a arrearage and undertax customer list, a personal credit score, and the like, and is not limited herein.
Specifically, the manner of obtaining the debt information of the target customer may be obtained from an associated system, and may be, but is not limited to, an asset preservation management platform, a personal loan information system, personal loan customer information, a customer collection record, and the like, and is not limited herein.
S102, performing index extraction on the debt information to obtain at least one target index information.
The target index information includes, but is not limited to, payment flow, loan information, target customer information, hasty records, consumption information, property information, external data, and the like, and is not limited herein.
S103, inputting all target index information into the collection prompting decision determining model, and outputting to obtain a collection prompting decision for a target client.
The collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models.
Optionally, in another embodiment of the present application, the method for constructing a pull-out decision determination model, as shown in fig. 2, specifically includes:
s201, constructing a training sample set.
The training sample set comprises historical debt information of at least one training sample client and the sensitivity of the training sample client aiming at each collection strategy.
Optionally, in another embodiment of the present application, an implementation manner of step S201, as shown in fig. 3, includes:
s301, obtaining historical debt information of the training sample client.
S302, index extraction is carried out on historical debt information of the training sample client to obtain at least one sample index information.
The specific implementation manners of steps S301 and S302 may refer to those in S101 and S102 in the above embodiments, which are not described herein again.
And S303, preprocessing each sample index information to obtain target preprocessing data.
Optionally, in another embodiment of the present application, an implementation manner of step S303, as shown in fig. 4, includes:
s401, for each sample index information, preprocessing structured data in the sample index information to obtain first preprocessed data.
Optionally, in another embodiment of the application, an implementation manner of step S401, as shown in fig. 5, includes:
s501, judging whether the missing value of the structured data is larger than a first threshold value or not for each piece of structured data.
The first threshold is set by a technician, and may be modified according to an actual application situation, which is not limited herein.
Specifically, missing value detection is performed on each piece of structured data to obtain a missing value of the structured data, and if it is determined that the missing value of the structured data is greater than the first threshold, step S502 is executed; if the missing value of the structured data is not greater than the first threshold, S503 is executed.
And S502, deleting the structured data.
S503, judging whether the structured data is a continuous variable.
Specifically, if the structured data is determined to be a continuous variable, step S504 is executed; if the structured data is not determined to be a continuous variable, step S506 is executed.
And S504, carrying out box separation on the structured data.
It should be noted that after the binning is performed, it is to be ensured that the structured data in each bin is greater than a certain percentage, and this percentage is also set by a skilled person and can be changed according to the actual application situation, and is not limited herein.
And S505, selecting an information value of the boxed structured data to obtain first preprocessing data.
The information value is an (information value) value, and is mainly used for encoding the input variable and estimating the prediction capability. And a threshold value can be set for the IV value, and this threshold value is also set by the technician, and can be changed according to the actual application, and is not limited herein.
S506, judging whether the level number of the structured data is larger than a second threshold value.
The second threshold is set by a technician, and may be modified according to an actual application situation, which is not limited herein.
Specifically, if the number of levels of the structured data is greater than the second threshold, step S507 is executed; if the level number of the structured data is not greater than the second threshold, step S505 is executed.
And S507, carrying out variable clustering on the structured data.
The variable clustering may adopt, but is not limited to, a chi-square selection method, and is not limited herein.
Optionally, in another embodiment of the present application, in an implementation manner of the method for determining a catalyst recovery decision, the method further includes:
all bins are optimized and structured data with co-linearity in each bin is removed.
Wherein, optimize the case to guarantee that the model of training can not be because of the case is careful, thereby lead to the problem of overfitting. And carrying out co-linearity check on each box, and removing the structured data with co-linearity in each box.
S402, preprocessing unstructured data in the sample index information aiming at each sample index information to obtain second preprocessed data.
Optionally, in another embodiment of the present application, an implementation manner of step S402, as shown in fig. 6, includes:
s601, performing corpus cleaning on the unstructured data to obtain first target data.
Specifically, a large amount of unstructured data may be collected, and then subjected to punctuation, numbers, english, special symbols, and the like by using a regular expression method, and cleaned in a null value table, a defect, and the like, which is not limited herein.
S602, performing word segmentation processing on the first target data to obtain second target data.
Specifically, the automatic Chinese word segmentation can be performed by using an open source module library and a word segmentation tool which are not limited to Chinese word segmentation, and the method is not limited herein.
And S603, extracting keywords from the second target data to obtain third target data.
Specifically, but not limited to, extracting the keywords by using word frequency may be adopted, and is not limited herein.
And S604, performing feature extraction on the third target data to obtain second preprocessing data.
It should be noted that, considering the vector economy of Word2Vec and the ability to express the close relationship (such as cosine similarity model) between words, the Word2Vec algorithm can be used to perform feature extraction. Similarly, other feature extraction methods may be used to perform feature extraction, and are not limited herein.
And S403, taking all the first preprocessed data and all the second preprocessed data as target preprocessed data.
S304, corresponding rules are formulated for each sample index information respectively, and corresponding collection handling strategies are set for each sample index group.
The sample index comprises at least one piece of sample index information. The collection handling policy includes, but is not limited to, collection with telephone, collection with going to the home, collection with going to the outsourcing, collection with judicial law, and payment due.
For example: making a corresponding rule for the index A, a corresponding rule for the index B, a corresponding rule for the index C and the like; corresponding collection and disposal strategies are formulated for the index A, the index B and the index C; making corresponding collection and disposal strategies for the index A and the index C; and establishing a corresponding collection and disposal strategy for the index A.
Suppose that: if index a is the mandatory execution time of the personal loan, a rule may be formulated for index a, such as: the obligation execution time of the personal loan exceeds N days and the debt can not be reclaimed.
S305, determining the sensitivity of the training sample client to each collection handling strategy.
The method mainly refers to the sensitivity of a client to an acceptance handling strategy, and reflects the effectiveness of applying a certain acceptance handling strategy to the client.
Specifically, since the collection-promoting disposal policies are not in an obvious sequence, the end of the previous collection-promoting disposal policy is declared as long as the next collection-promoting disposal policy is met. Based on this, observation period windows for different revenue-forcing treatment strategies may be defined. In order to investigate the sensitivity of a training sample client to an acceptance handling strategy, when the acceptance handling strategy is obtained and adopted again, the present value of the loan application recovery amount of the acceptance handling strategy accounts for the proportion of the owed loan interest, wherein the daily loan execution interest rate is selected as the discount rate during the present value calculation, and the sensitivity Y calculation formula of the client acceptance handling strategy is as follows:
wherein T is the time from the default date in days.
S202, inputting historical debt information of the training sample client into a plurality of two-classification prediction models in a preset model to obtain a prediction result aiming at each collection urging strategy.
The multi-classification task can be decomposed into a plurality of two-classification tasks for solving, and finally the prediction results output by the two-classification prediction models are the prediction results obtained by fusing all the sub-models in a majority voting mode. For example: and 5 binary prediction models are provided, wherein 3 binary prediction models are effective to the prediction result of call incoming, and 2 binary prediction models are ineffective to the prediction result of call incoming, so that the final obtained result is effective to the prediction result of call incoming.
It should be noted that, a plurality of binary prediction models can obtain initial model prediction probabilities, and the probability scores can be filtered according to the number of days with bad entering. For example, if the number of bad days for a debt item is more than 30 days, the call earning probability is decreased, which is not limited herein.
S203, aiming at each collection strategy, predicting the effectiveness of the collection strategy by using an intelligent decision model in a preset model to obtain a first prediction result.
Wherein, the intelligent decision model is a logistic regression model.
Specifically, according to historical debt information of a training sample client and a prediction result of an induced collection strategy output from a plurality of two-classification prediction models, the effectiveness of the induced collection strategy is predicted, and a first prediction result is obtained.
And S204, calibrating the first prediction result by using an expert rule model in a preset model to obtain a second prediction result.
Wherein, the expert rule model is a preset rule model.
It should be noted that, if the number of samples is small, the method is more suitable for direct prediction by using an expert rule model.
It should be noted that, filtering may be performed according to a rule hit result, but is not limited to, if a certain debt item hits a certain expert rule in the expert rule model, the corresponding probability score is increased, and this is not limited herein.
After filtering according to rule hits, filtering may also be through a list. For example: and adjusting the corresponding probability score by matching the list of the poor subscribers with the established card, the blacklist of the debt and the like.
After multiple rounds of filtering, the probability score may be out of the range of 0-1, and for this, we map the probability score by using a softmax function. Suppose we have an array V, ViRepresents the ith score in V; p is a radical ofiThe score is adjusted by three rounds; probiMeans the converted score; j is a circular index, and traverses all scores, then the softmax value of this score is:
according to the characteristic of softmax, we can map probability values into a range of (0, 1), and guarantee that the sum of the probabilities is 1.
S205, judging whether the second prediction result of the collection urging strategy and the error of the sensitivity of the training sample client for the collection urging strategy meet the preset convergence condition.
Specifically, if the error between the second prediction result of the collection forcing strategy and the sensitivity of the training sample client for the collection forcing strategy is determined to satisfy the preset convergence condition, the step S206 is executed; if the error between the second prediction result of the collection forcing strategy and the sensitivity of the training sample client to the collection forcing strategy is determined not to satisfy the preset convergence condition, step S207 is executed.
S206, taking the preset model as a collection decision determination model.
And S207, adjusting a preset model by using the second prediction result of the collection urging strategy and the error of the training sample client aiming at the sensitivity of the collection urging strategy.
It should be noted that, in the actual application process, relevant information is presented to business personnel through the front-end interface. The front end interface mainly comprises: and (5) inquiring a box. And provides the company, the client name, the loan account number, the loan type, the priority recommendation strategy and the like for the business personnel to quickly inquire; basic information of bad credits. Covering an operator, a client number, a client name, a loan account number, a product name, a loan type, bad entering time, balance and the like; and preferentially recommending the strategy. Displaying two types of collection-hastening disposal strategies with highest effectiveness to be presented to business personnel; a validity probability value. And displaying the effectiveness probability score of the collection handling strategy. The model results are run and batched daily, and are pushed to the service platform according to the time efficiency of T + 2. For the data in the current month, the adopted data storage strategy is to keep the data by day; for historical data of an improper month, the data storage strategy can automatically archive the historical data.
According to the scheme, the method for determining the hasty decision comprises the following steps: firstly, acquiring debt information of a target client; then, performing index extraction on the debt information to obtain at least one target index information; finally, inputting all target index information into a collection prompting decision determining model, and outputting to obtain a collection prompting decision for a target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models. So as to achieve the purpose of quickly and accurately determining the hastening decision aiming at the target client.
Another embodiment of the present application provides a device for determining a hasty decision, as shown in fig. 7, specifically including:
a first obtaining unit 701, configured to obtain debt information of a target customer.
A first extracting unit 702, configured to perform index extraction on the debt information to obtain at least one target index information.
The first determining unit 703 is configured to input all target index information into the collection decision determining model, and output a collection decision for the target customer.
The collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models.
For a specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, as shown in fig. 1, which is not described herein again.
Optionally, in another embodiment of the application, an implementation of the pull-up decision determination model building unit includes:
and the training sample set constructing unit is used for constructing a training sample set.
The training sample set comprises historical debt information of at least one training sample client and the sensitivity of the training sample client aiming at each collection strategy.
And the input unit is used for inputting the historical debt information of the training sample client into a plurality of two-classification prediction models in a preset model to obtain a prediction result aiming at each collection urging strategy.
And the prediction unit is used for predicting the effectiveness of the collection strategies by utilizing an intelligent decision model in a preset model aiming at each collection strategy to obtain a first prediction result.
Wherein, the intelligent decision model is a logistic regression model.
And the calibration unit is used for calibrating the first prediction result by utilizing an expert rule model in a preset model to obtain a second prediction result.
Wherein, the expert rule model is a preset rule model.
And the adjusting unit is used for adjusting the preset model by utilizing the second prediction result of the collection prompting strategy and the error of the training sample client on the sensitivity of the collection prompting strategy until the error of the second prediction result of the adjusted preset model output on the collection prompting strategy and the error of the training sample client on the sensitivity of the collection prompting strategy meet the preset convergence condition, and taking the adjusted preset model as a collection prompting decision determining model.
For a specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, as shown in fig. 2, which is not described herein again.
Optionally, in another embodiment of the present application, an implementation manner of the sample set constructing unit includes:
and the second acquisition unit is used for acquiring the historical debt information of the training sample client.
And the second extraction unit is used for performing index extraction on the historical debt information of the training sample client to obtain at least one sample index information.
And the preprocessing unit is used for preprocessing each sample index information to obtain target preprocessing data.
And the setting unit is used for respectively setting corresponding rules for each sample index information and setting corresponding collection and disposal strategies for each sample index group.
Wherein the set of sample metrics includes at least one of the sample metric information.
And the second determination unit is used for determining the sensitivity of the training sample client to each induced charge treatment strategy.
For a specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, as shown in fig. 3, which is not described herein again.
Optionally, in another embodiment of the present application, an implementation manner of the preprocessing unit includes:
and the first preprocessing subunit is used for preprocessing the structured data in the sample index information aiming at each sample index information to obtain first preprocessed data.
And the second preprocessing subunit is used for preprocessing the unstructured data in the sample index information according to each piece of sample index information to obtain second preprocessed data.
And the third determining unit is used for taking all the first preprocessing data and all the second preprocessing data as target preprocessing data.
For a specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, as shown in fig. 4, which is not described herein again.
Optionally, in another embodiment of the present application, an implementation manner of the first preprocessing subunit includes:
the first judging unit is used for judging whether the missing value of the structured data is larger than a first threshold value or not aiming at each piece of structured data.
And the deleting unit deletes the structured data if the first judging unit judges that the missing value of the structured data is greater than the first threshold value.
And the second judging unit is used for judging whether the structured data is a continuous variable or not if the first judging unit judges that the missing value of the structured data is not larger than the first threshold value.
And the box dividing unit is used for dividing the structured data into boxes if the structured data is judged to be the continuous variable by the second judging unit.
And the information value selection unit is used for selecting the information value of the classified structured data to obtain first preprocessing data.
And a third judging unit configured to judge whether the level number of the structured data is greater than the second threshold value if the second judging unit judges that the structured data is not the continuous variable.
And the variable clustering unit is used for performing variable clustering on the structured data if the third judging unit judges that the level number of the structured data is greater than the second threshold value.
And the information value selection unit is also used for selecting the information value of the structured data after variable clustering to obtain first preprocessing data.
And the information value selecting unit is further used for selecting the information value of the structured data to obtain the first preprocessed data if the third judging unit judges that the level number of the structured data is not greater than the second threshold value.
For a specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, as shown in fig. 5, which is not described herein again.
Optionally, in another embodiment of the present application, an implementation manner of the decision prompting device further includes:
and the optimization unit is used for optimizing all the sub-boxes.
And the removing unit is used for removing the structured data with the colinearity in each box.
For specific working processes of the units disclosed in the above embodiments of the present application, reference may be made to the contents of the corresponding method embodiments, which are not described herein again.
Optionally, in another embodiment of the present application, an implementation manner of the second preprocessing subunit includes:
and the cleaning unit is used for obtaining the first target data after the corpus cleaning is carried out on the unstructured data.
And the word segmentation unit is used for carrying out word segmentation processing on the first target data to obtain second target data.
And the keyword extraction unit is used for extracting keywords from the second target data to obtain third target data.
And the feature extraction unit is used for extracting features of the third target data to obtain second preprocessing data.
For a specific working process of the unit disclosed in the above embodiment of the present application, reference may be made to the content of the corresponding method embodiment, as shown in fig. 6, which is not described herein again.
According to the above scheme, the device for determining the hasty check-out decision provided by the present application comprises: first, a first obtaining unit 701 obtains debt information of a target client; then, the first extraction unit 702 performs index extraction on the debt information to obtain at least one target index information; finally, the first determining unit 703 inputs all target index information into the collection decision determining model, and outputs the information to obtain a collection decision for the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models. So as to achieve the purpose of quickly and accurately determining the hastening decision aiming at the target client.
Another embodiment of the present application provides a server, as shown in fig. 8, including:
one or more processors 801.
A storage device 802 on which one or more programs are stored.
The one or more programs, when executed by the one or more processes 801, cause the one or more processors 801 to implement a method of determining an hasten decision as described in any of the above embodiments.
Another embodiment of the present application provides a computer storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the method for determining an hasten decision as described in any one of the above embodiments.
In the above embodiments disclosed in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The apparatus and method embodiments described above are illustrative only, as the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules in the embodiments of the present disclosure may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part. The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present disclosure may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a live broadcast device, or a network device) to execute all or part of the steps of the method according to the embodiments of the present disclosure. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Those skilled in the art can make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for determining a catalyst recovery decision, comprising:
acquiring debt information of a target client;
performing index extraction on the debt information to obtain at least one target index information;
inputting all the target index information into a collection prompting decision determining model, and outputting to obtain a collection prompting decision for the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models.
2. The method of determining according to claim 1, wherein the pull-up decision determination model is constructed by a method comprising:
constructing a training sample set; wherein the training sample set comprises historical debt information of at least one training sample client and the sensitivity of the training sample client for each charging strategy;
inputting the historical debt information of the training sample client into a plurality of two-classification prediction models in a preset model to obtain a prediction result aiming at each collection strategy;
for each collection strategy, predicting the effectiveness of the collection strategy by using an intelligent decision model in the preset model to obtain a first prediction result; the intelligent decision model is a logistic regression model;
calibrating the first prediction result by using an expert rule model in the preset model to obtain a second prediction result; the expert rule model is a preset rule model;
and adjusting the preset model by utilizing the second prediction result of the collection prompting strategy and the error of the training sample client for the sensitivity of the collection prompting strategy until the error of the adjusted preset model output for the second prediction result of the collection prompting strategy and the error of the training sample client for the sensitivity of the collection prompting strategy meet the preset convergence condition, and taking the adjusted preset model as a collection prompting decision determination model.
3. The method of claim 2, wherein the constructing the training sample set comprises:
acquiring historical debt information of a training sample client;
index extraction is carried out on the historical debt information of the training sample client to obtain at least one sample index information;
preprocessing each sample index information to obtain target preprocessing data;
respectively formulating a corresponding rule for each sample index information, and setting a corresponding collection and disposal strategy for each sample index group; wherein the set of sample metrics includes at least one of the sample metric information;
determining a sensitivity of a training sample customer to each of the revenue handling policies.
4. The method of claim 3, wherein the preprocessing each of the sample metric information to obtain target preprocessed data comprises:
for each sample index information, preprocessing structured data in the sample index information to obtain first preprocessed data;
for each piece of sample index information, preprocessing unstructured data in the sample index information to obtain second preprocessed data;
and taking all the first preprocessing data and all the second preprocessing data as target preprocessing data.
5. The determination method according to claim 4, wherein the preprocessing the structured data in the sample index information to obtain first preprocessed data for each sample index information comprises:
for each piece of the structured data, judging whether the missing value of the structured data is larger than a first threshold value;
if the missing value of the structured data is judged to be larger than a first threshold value, deleting the structured data;
if the missing value of the structured data is judged to be not larger than a first threshold value, judging whether the structured data is a continuous variable;
if the structured data are judged to be continuous variables, carrying out box separation on the structured data;
selecting information values of the boxed structured data to obtain first preprocessing data;
if the structured data is judged not to be a continuous variable, judging whether the level number of the structured data is larger than a second threshold value;
if the level number of the structured data is judged to be larger than a second threshold value, performing variable clustering on the structured data;
selecting information values of the structured data after variable clustering to obtain first preprocessing data;
and if the level number of the structured data is judged to be not larger than a second threshold value, selecting an information value of the structured data to obtain first preprocessing data.
6. The determination method according to claim 5, further comprising:
optimizing all the bins;
removing the structured data with co-linearity in each of the bins.
7. The determination method according to claim 4, wherein the preprocessing the unstructured data in the sample index information for each sample index information to obtain second preprocessed data comprises:
performing corpus cleaning on the unstructured data to obtain first target data;
performing word segmentation processing on the first target data to obtain second target data;
extracting keywords from the second target data to obtain third target data;
and performing feature extraction on the third target data to obtain second preprocessing data.
8. An apparatus for determining a catalyst decision, comprising:
the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring debt information of a target client;
the first extraction unit is used for carrying out index extraction on the debt information to obtain at least one target index information;
the first determining unit is used for inputting all the target index information into a collection decision determining model and outputting the collection decision aiming at the target client; the collection decision determining model consists of an intelligent decision model, an expert rule model and a plurality of two classification models.
9. A server, comprising:
one or more processors;
a storage device having one or more programs stored thereon;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method for determining an hasten decision of any of claims 1-7.
10. A computer storage medium, having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method of determining an hasten decision of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110494902.5A CN113112347A (en) | 2021-05-07 | 2021-05-07 | Determination method of hasty collection decision, related device and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110494902.5A CN113112347A (en) | 2021-05-07 | 2021-05-07 | Determination method of hasty collection decision, related device and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113112347A true CN113112347A (en) | 2021-07-13 |
Family
ID=76721230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110494902.5A Pending CN113112347A (en) | 2021-05-07 | 2021-05-07 | Determination method of hasty collection decision, related device and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113112347A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113657901A (en) * | 2021-07-23 | 2021-11-16 | 上海钧正网络科技有限公司 | Method, system, terminal and medium for managing collection of owing user |
| CN113822490A (en) * | 2021-09-29 | 2021-12-21 | 平安银行股份有限公司 | Asset clearing and accepting method and device based on artificial intelligence and electronic equipment |
-
2021
- 2021-05-07 CN CN202110494902.5A patent/CN113112347A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113657901A (en) * | 2021-07-23 | 2021-11-16 | 上海钧正网络科技有限公司 | Method, system, terminal and medium for managing collection of owing user |
| CN113657901B (en) * | 2021-07-23 | 2024-04-16 | 上海钧正网络科技有限公司 | Method, system, terminal and medium for managing fee owed users |
| CN113822490A (en) * | 2021-09-29 | 2021-12-21 | 平安银行股份有限公司 | Asset clearing and accepting method and device based on artificial intelligence and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2017133456A1 (en) | Method and device for determining risk evaluation parameter | |
| US20120136812A1 (en) | Method and system for machine-learning based optimization and customization of document similarities calculation | |
| CN111368147B (en) | Graph feature processing method and device | |
| CN113112347A (en) | Determination method of hasty collection decision, related device and computer storage medium | |
| CN109583731B (en) | Risk identification method, device and equipment | |
| CN110796485A (en) | Method and device for improving prediction precision of prediction model | |
| CN112070615A (en) | Financial product recommendation method and device based on knowledge graph | |
| CN111967807A (en) | Method and device for generating risk event judgment rule executed by computer | |
| CN113177839A (en) | Credit risk assessment method, device, storage medium and equipment | |
| CN111259922A (en) | Order data processing method and device based on customer order-returning early warning | |
| CN108197795B (en) | Malicious group account identification method, device, terminal and storage medium | |
| WO2020024444A1 (en) | Group performance grade recognition method and apparatus, and storage medium and computer device | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN114239697A (en) | Target object classification method and device, electronic equipment and storage medium | |
| CN111435357A (en) | Gas station customer churn prediction method, gas station customer churn prediction device, electronic equipment and storage medium | |
| CN112801784A (en) | Bit currency address mining method and device for digital currency exchange | |
| CN111951008A (en) | Risk prediction method and device, electronic equipment and readable storage medium | |
| CN114841705B (en) | Anti-fraud monitoring method based on scene recognition | |
| CN111143533A (en) | Customer service method and system based on user behavior data | |
| CN115965468A (en) | Transaction data-based abnormal behavior detection method, device, equipment and medium | |
| CN114841588A (en) | Information processing method, device, electronic equipment and computer readable medium | |
| CN114756685A (en) | Complaint risk identification method and device for complaint sheet | |
| CN113065892A (en) | Information pushing method, device, equipment and storage medium | |
| CN112200488A (en) | Risk identification model training method and device for business object | |
| CN113128217B (en) | Public opinion disposition decision-making method based on network twinning space |
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 |