CN111581252A - Dynamic collection urging method and system based on multi-dimensional information data - Google Patents
Dynamic collection urging method and system based on multi-dimensional information data Download PDFInfo
- Publication number
- CN111581252A CN111581252A CN202010369251.2A CN202010369251A CN111581252A CN 111581252 A CN111581252 A CN 111581252A CN 202010369251 A CN202010369251 A CN 202010369251A CN 111581252 A CN111581252 A CN 111581252A
- Authority
- CN
- China
- Prior art keywords
- data
- collection
- information data
- variable
- value
- 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
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 82
- 238000012216 screening Methods 0.000 claims abstract description 9
- 239000013074 reference sample Substances 0.000 claims description 26
- 230000002159 abnormal effect Effects 0.000 claims description 15
- 230000003197 catalytic effect Effects 0.000 claims description 14
- 238000007670 refining Methods 0.000 claims description 14
- 238000011084 recovery Methods 0.000 claims description 12
- 239000000523 sample Substances 0.000 claims description 10
- 230000004931 aggregating effect Effects 0.000 claims description 8
- 235000019580 granularity Nutrition 0.000 claims description 8
- 238000003306 harvesting Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 5
- 238000004364 calculation method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000009286 beneficial effect Effects 0.000 description 5
- 230000001133 acceleration Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 4
- 230000003203 everyday effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000000945 filler Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000001737 promoting effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000006555 catalytic reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- 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
Abstract
The invention discloses a dynamic collection method and a system based on multi-dimensional information data, wherein the method comprises the following steps: s1 is used for dynamically capturing dynamic collection target; s2 is used for acquiring multi-dimensional information data of the collection object; s3 is used for carrying out data processing on the multidimensional information data of the collection object; s4 is used for carrying out variable screening on the processed multidimensional information data; and S5 is used for carrying out scoring prediction on the screened variables and carrying out dynamic collection on the automatically judged risk of the client. The invention provides a dynamic collection method and a system based on multi-dimensional information data, and aims to dynamically score customers by using the multi-dimensional information data. According to the result of the collection urging scoring, the risk of the client is automatically judged, and a collector and a collection urging strategy corresponding to the client are reasonably determined, so that the collection urging effect is favorably improved, the bad account loss can be reduced, and flexible collection urging management is realized.
Description
Technical Field
The invention relates to the technical field of computer software, in particular to a dynamic collection urging method and system based on multi-dimensional information data.
Background
The characteristics of large number of strokes and small amount of money in a single credit business determine the importance of using technical means in post-credit collection management. In recent years, loan industries such as consumer finance, low-volume loan, P2P, and the like have been developed. Meanwhile, the domestic credit investigation system has a plurality of defects and lacks of a metering tool for subdividing customers. In the existing collection method, customers are usually distinguished according to the length of overdue time, so that the collection promotion fine management degree is not high, and high-risk customers which become bad customers and low-risk customers which can pay for money actively cannot be distinguished in an early stage. The former is not subjected to a strong receiving expediting method as soon as possible, so that the former is changed into a bad client; the latter is excessively hastened, so that overdue bad account rate is high, and unnecessary recovery cost is increased. In addition, different harvest promoting strategies are adopted, and different harvest promoting effects can be generated.
At present, although there are cases of using the collection urging score to control the bad assets in the industry, the collection urging score is mainly developed by researching the behaviors related to the historical repayment of the client, and the historical collection urging record of the client and the collection urging effect generated by the record urging record are not deeply mined. Furthermore, existing revenue collection methods do not combine the credit score before the client is credited to perform multidimensional modeling on the client. In addition, most of the collection scores in the industry are static, updating and iteration can not be performed every day, and dynamic scores can not be formed, so that effective post-loan management can not be performed on clients more accurately.
Disclosure of Invention
The invention aims to provide a dynamic collection method and a system based on multi-dimensional information data.
The invention provides a dynamic collection method based on multi-dimensional information data, which comprises the following steps: s1 is used for dynamically capturing dynamic collection target; s2 is used for acquiring multi-dimensional information data of the collection object; s3 is used for carrying out data processing on the multidimensional information data of the collection object; s4 is used for carrying out variable screening on the processed multidimensional information data; and S5 is used for carrying out score prediction on the screened variables and carrying out dynamic collection on different customers.
The step of S2 for acquiring multidimensional information data of the collection target includes: s21, acquiring original data of different-dimension collection targets; and S22 is used for processing and refining the collected original data. The step of S22, which is used to refine the collected raw data, includes: s221 is used for counting different parameter information of the information event according to the time interval and outputting the information value; s222, counting different parameter information of the information event according to different client states and outputting an information value; and S223, aggregating the information values to different statistical granularities for numerical processing. The step of S3, which is to perform data processing on the multidimensional information data of the collection target, includes: s31 is used for processing extreme values or abnormal values in the multi-dimensional information data; s32 is a step for processing missing values in the multidimensional information data; s33 is a step of processing the categorical variable data in the multidimensional information data and replacing each category of the categorical variable with a numerical value. The step of S31 for processing extreme values or abnormal values in the multidimensional information data includes: s311 is used for determining an upper limit value and a lower limit value of the catalytic recovery target variable; s312, when the value of each catalytic target variable is taken, if the value is larger than the upper bound value, the value is re-assigned as the upper bound value; if the numerical value is smaller than the lower bound value, the numerical value is reassigned as the lower bound value; otherwise, the steps are unchanged. The step of S32 for processing missing values in the multi-dimensional information data includes: s321, when processing numerical variable data, filling missing values in each variable by an average value of the variable; s322 is a step of filling missing values in each variable with the mode of the variable when processing the variable data of the type. The step of S5, which is used to predict the score of the screened variables and dynamically urge to accept the risk of the customer by automatic judgment, includes: s51 is used for dividing the data of the screened variable into a reference sample and a non-reference sample; s52 is used for judging whether the screened variables are reference samples or not, and if the screened variables are the reference samples, the customer data rules of the reference samples are obtained through an algorithm and the prediction and scoring steps are carried out; if the sample is a non-reference sample, performing prediction and scoring; and S53 is used for segmenting and dividing the client according to the prediction score and performing dynamic collection.
The invention provides a dynamic collection system based on multi-dimensional information data, which comprises: a module for dynamically capturing dynamic collection targets; the module is used for acquiring multi-dimensional information data of the collection target; the module is used for carrying out data processing on the multidimensional information data of the collection object; the module is used for carrying out variable screening on the processed multidimensional information data; and the module is used for grading and predicting the screened variables and carrying out dynamic collection on different clients.
The module for acquiring the multidimensional information data of the collection target comprises: a submodule for collecting original data of different-dimension collection targets; and the submodule is used for processing and refining the collected original data. The submodule for processing and refining the collected original data comprises: a unit for counting different parameter information of information event according to time interval and outputting information value; a unit for counting different parameter information of information event according to different client states and outputting information value; and the unit is used for aggregating the information values to different statistical granularities for numerical processing. The module for processing the multidimensional information data of the collection object comprises: the submodule is used for processing extreme values or abnormal values in the multi-dimensional information data; the submodule is used for processing missing values in the multi-dimensional information data; and the submodule is used for processing the category type variable data in the multi-dimensional information data and replacing each category in the category type variable with a numerical value. The submodule for processing extreme values or abnormal values in the multi-dimensional information data comprises: the unit is used for determining an upper limit value and a lower limit value of the catalytic recovery target variable; when each catalytic recovery target variable is valued, if the value is greater than the upper bound value, the value is re-assigned as the upper bound value; if the numerical value is smaller than the lower bound value, the numerical value is reassigned as the lower bound value; and the other cases are unchanged. The sub-module for processing missing values in the multi-dimensional information data comprises: a unit for filling missing values in each variable by an average value of the variable when processing numerical variable data; and a unit for filling missing values in each variable by the mode of the variable when processing variable data of the type. The module for scoring and predicting the screened variables and dynamically urging to automatically judge the risk of the client comprises: a submodule for dividing the data of the screened variables into reference samples and non-reference samples; the step of judging whether the screened variables are reference samples or not, and if the variables are the reference samples, acquiring the customer data rules of the reference samples through an algorithm and predicting and scoring; if the sample is a non-reference sample, performing prediction scoring; and the submodule is used for segmenting the clients according to the prediction scores and performing dynamic collection.
The invention provides a dynamic collection method and a system based on multi-dimensional information data, and aims to dynamically score customers by using the multi-dimensional information data. And automatically judging the risk of the client according to the collection prompting grading result, and reasonably determining a collector prompting and a collection prompting strategy corresponding to the client. Not only is beneficial to improving the effect of collection hastening, but also can reduce the bad account loss. The method for urging collection is developed based on multi-dimensional characteristics such as historical repayment information, historical urging collection behaviors and credit scores before credit, and the customer urging collection scores are dynamic. Therefore, business personnel can change the collection urging strategy at any time along with the change of the credit environment or the collection urging target population, and flexible collection urging management is realized.
Drawings
Fig. 1 and 2 are schematic diagrams illustrating steps of a dynamic collection method based on multidimensional information data according to an embodiment of the present invention;
fig. 3 is a schematic diagram of the step of S2 for acquiring multidimensional information data of an object to be induced to receive according to an embodiment of the present invention;
fig. 4 is a schematic diagram of the step S22 of refining the collected raw data according to the first embodiment of the present invention;
fig. 5 is a schematic diagram of the step of S3 for performing data processing on the multidimensional information data of the catalysis target according to the first embodiment of the present invention;
fig. 6 is a schematic diagram of the processing steps performed by S31 for extreme values or abnormal values in the multi-dimensional information data according to the first embodiment of the present invention;
fig. 7 is a schematic diagram of the step of S32 for processing the missing value in the multi-dimensional information data according to the first embodiment of the present invention;
fig. 8 is a schematic diagram of the step of S5, according to an embodiment of the present invention, for performing score prediction on the screened variables and performing dynamic revenue generation on the automatically determined risk of the customer.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, 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 some, but not all, embodiments of the present invention. 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.
Example one
As shown in fig. 1 and 2, the present embodiment provides a dynamic collection method based on multidimensional information data, including the following steps:
s1 is used for dynamically capturing dynamic collection target;
s2 is used for acquiring multi-dimensional information data of the collection object;
s3 is used for carrying out data processing on the multidimensional information data of the collection object;
s4 is used for carrying out variable screening on the processed multidimensional information data;
and S5 is used for carrying out scoring prediction on the screened variables and carrying out dynamic collection on the automatically judged risk of the client.
As can be understood by those skilled in the art, the invention mainly aims to provide a hastening method based on multi-dimensional data information, aiming at dynamically scoring customers by utilizing the multi-dimensional data information. And automatically judging the risk of the client according to the collection prompting grading result, and reasonably determining a collector prompting and a collection prompting strategy corresponding to the client. Not only is beneficial to improving the effect of collection hastening, but also can reduce the bad account loss. The method for urging collection is developed based on multi-dimensional characteristics such as historical repayment information, historical urging collection behaviors and credit scores before credit, and the customer urging collection scores are dynamic. Therefore, business personnel can change the collection urging strategy at any time along with the change of the credit environment or the collection urging target population, and flexible collection urging management is realized.
And acquiring the collection target by using the dynamic catcher. The dynamic catcher dynamically screens out the clients needing to hasten or adjust hastening strategies from all unclosed cases every day, namely hastening targets. The collection target comprises a client who has a new bill and is overdue on the same day and a client who has not paid for the full amount from the previous monthly bill date to the beginning of the month, namely, each client can be marked as the collection target by the dynamic catcher for a plurality of times.
As shown in fig. 3, the step S2 of acquiring multidimensional information data of an object to be induced to receive includes:
s21, acquiring original data of different-dimension collection targets;
and S22 is used for processing and refining the collected original data.
As will be appreciated by those skilled in the art, the data collector is utilized to obtain multidimensional information data for the collection-forcing target. The data collector comprises ten-dimensional collection, processing and extraction processes of overdue information, short message information, current bill information, historical bill information, communication information, credit scoring information, historical collection information, historical repayment information, historical credit clearing information, basic information and the like of a collection target, wherein each dimension comprises a certain number of variables. The raw data is the data collected from the different dimensions and not processed.
As shown in fig. 4, the step S22 of refining the collected raw data includes:
s221 is used for counting different parameter information of the information event according to the time interval and outputting the information value;
s222, counting different parameter information of the information event according to different client states and outputting an information value;
and S223, aggregating the information values to different statistical granularities for numerical processing.
Those skilled in the art can understand that the parameter information includes the occurrence frequency, duration and amount of the above information events, and the numerical processing mode includes addition, maximum value calculation, minimum value calculation and the like; the processing and refining process refers to a process of carrying out secondary construction on the collected original data, and comprises the steps of counting the times, duration and amount of occurrence of the information events within a fixed time interval from the present, counting the times, duration and amount of occurrence of the information events according to the state of a client, aggregating the information values to different statistical granularities, and carrying out addition, maximum value calculation, minimum value calculation and the like.
As shown in fig. 5, the step S3 of performing data processing on the multidimensional information data of the collection target includes:
s31 is used for processing extreme values or abnormal values in the multi-dimensional information data;
s32 is a step for processing missing values in the multidimensional information data;
s33 is a step of processing the categorical variable data in the multidimensional information data and replacing each category of the categorical variable with a numerical value.
As shown in fig. 6, the step S31 of processing extreme values or abnormal values in the multidimensional information data includes:
s311 is used for determining an upper limit value and a lower limit value of the catalytic recovery target variable;
s312, when the value of each catalytic target variable is taken, if the value is larger than the upper bound value, the value is re-assigned as the upper bound value; if the numerical value is smaller than the lower bound value, the numerical value is reassigned as the lower bound value; otherwise, the steps are unchanged.
As shown in fig. 7, the step S32 of processing missing values in the multi-dimensional information data includes:
s321, when processing numerical variable data, filling missing values in each variable by an average value of the variable;
s322 is a step of filling missing values in each variable with the mode of the variable when processing the variable data of the type.
As can be understood by those skilled in the art, the embodiment utilizes an exception handler, a filler and an encoder in the system to perform data processing on the multidimensional information data of the collection object. The exception handler comprises a process for processing extreme values and abnormal values in the multi-dimensional information data, and the main steps are as follows: for a variable to be processed, determining an upper bound value and a lower bound value of the variable; and aiming at the value of each variable of the collection target, if the value is greater than the upper bound value, the variable is reassigned as the upper bound value, if the value is less than the lower bound value, the variable is reassigned as the lower bound value, and the other conditions are not changed. The stuffer processes missing values in the multi-dimensional information data, and the main steps are as follows: for numerical variable data, the missing value in each variable is filled with the average value of the variable; for categorical variable data, the missing value in each variable is filled with the mode of the variable. The encoder processes the categorical variable data in the multi-dimensional information data, and replaces each category in the categorical variable with a numerical value, such as "male" equal to 1 and "female" equal to 2.
And performing variable screening on the processed multi-dimensional information data by using a variable screener, and rejecting variables with low payment relevance between each dimension and the client and variables with strong correlation between each dimension and relatively low payment relevance between the same dimension and the client.
As shown in fig. 8, the step S5 of performing score prediction on the filtered variables and dynamically hastening revenue for automatically judging the risk of the customer includes:
s51 is used for dividing the data of the screened variable into a reference sample and a non-reference sample;
s52 is used for judging whether the screened variables are reference samples or not, and if the screened variables are the reference samples, the customer data rules of the reference samples are obtained through an algorithm and the prediction and scoring steps are carried out; if the sample is a non-reference sample, performing prediction and scoring;
and S53 is used for segmenting and dividing the client according to the prediction score and performing dynamic collection.
One skilled in the art will appreciate that the data for the screened variables described above are divided into reference samples and non-reference samples. The reference sample is the customer data with the collection record and payment label, and is used for training the model. The non-reference sample is data which has no payment label and needs to evaluate the payment willingness of the customer, namely the latest batch of data related to the customer who needs to implement the charging method. And if the sample is the reference sample, putting the reference sample into a learner to obtain a prediction scorer. The learner acquires data rules of good and bad clients in the reference sample through a machine learning algorithm, the data rules comprise data characteristics of the clients which are biased to repayment in all dimensions, data characteristics of the clients which are biased to overdue in all dimensions, frequency distribution of the good and bad clients which are calibrated as collection-promoting targets by the dynamic capturer and the like, and the data rules are formed into a prediction scoring device by the learner. If the sample is a non-reference sample, the non-reference sample is directly scored by the prediction scorer, a client with a high repayment willingness is biased to get a high score without passing through a learner, and a client which is easy to overdue is biased to get a low score. Aiming at the customers with different scores, different collection urging strategies are implemented, and for the customers with high section, namely the customers with strong repayment willingness, the strategies of short-term low-frequency intelligent voice call collection urging or low-frequency short message reminding and later-period low-frequency artificial telephone collection urging are adopted; for low-segment customers, namely customers who are easy to be overdue, the strategies of short-term high-frequency intelligent voice call collection or high-frequency short message reminding and later-term high-frequency manual call collection are adopted.
The collection hastening method provided by the embodiment is beneficial to discovering high-risk customers as early as possible and reducing bad account loss; unifying the collection strategy standard; the catalytic recovery efficiency is improved. The high-risk customers predicted by the collection-urging score are listed as key collection-urging targets, and collection-urging force is increased by adopting collection-urging modes such as increasing collection-urging frequency and getting home in advance, so that the customers can be promoted to pay as soon as possible. And (4) carrying out judicial collection or ex-commission collection on overdue clients without repayment willingness among high-risk clients as soon as possible. The customer with different risks is distinguished by the collection grading, and then the customer is assigned to different types of collectors by the system, so that the consistency, objectivity and unbiased property of the collection mode and the risk characteristics of the customer are ensured. Meanwhile, the risk preference is adjusted at any time according to the grading threshold value. When the economic cycle or credit cycle changes, the receiving acceleration can be increased in time by adjusting the system, and the unified regulation and control of the receiving acceleration strategy can be realized. The personal credit service is characterized by large service volume and small single amount. The collection prompting method adopts a mode of automatically sending short messages and letters or making voice calls to low-risk customers to prompt the customers to pay on the premise of controllable risk, and does not prompt the customers who can actively pay to collect in the early stage. The collection urging method realizes the automatic management of personal credit business, improves the collection urging efficiency and reduces the operation cost and the labor cost. Meanwhile, the customer satisfaction is correspondingly improved because the high-quality customers are not disturbed.
Example two
The embodiment provides a dynamic collection system based on multi-dimensional information data, which includes:
a module for dynamically capturing dynamic collection targets;
the module is used for acquiring multi-dimensional information data of the collection target;
the module is used for carrying out data processing on the multidimensional information data of the collection object;
the module is used for carrying out variable screening on the processed multidimensional information data;
and the module is used for grading and predicting the screened variables and dynamically urging to automatically judge the risk of the client.
As can be appreciated by those skilled in the art, the primary objective of the present invention is to provide a system for soliciting revenue based on multidimensional data information, which is intended to dynamically score customers using multidimensional data information. And automatically judging the risk of the client according to the collection prompting grading result, and reasonably determining a collector prompting and a collection prompting strategy corresponding to the client. Not only is beneficial to improving the effect of collection hastening, but also can reduce the bad account loss. The method for urging collection is developed based on multi-dimensional characteristics such as historical repayment information, historical urging collection behaviors and credit scores before credit, and the customer urging collection scores are dynamic. Therefore, business personnel can change the collection urging strategy at any time along with the change of the credit environment or the collection urging target population, and flexible collection urging management is realized.
And acquiring the collection target by using the dynamic catcher. The dynamic catcher dynamically screens out the clients needing to hasten or adjust hastening strategies from all unclosed cases every day, namely hastening targets. The collection target comprises a client who has a new bill and is overdue on the same day and a client who has not paid for the full amount from the previous monthly bill date to the beginning of the month, namely, each client can be marked as the collection target by the dynamic catcher for a plurality of times.
Further, the module for acquiring multidimensional information data of the collection target comprises:
a submodule for collecting original data of different-dimension collection targets;
and the submodule is used for processing and refining the collected original data.
As will be appreciated by those skilled in the art, the data collector is utilized to obtain multidimensional information data for the collection-forcing target. The data collector comprises ten-dimensional collection, processing and extraction processes of overdue information, short message information, current bill information, historical bill information, communication information, credit scoring information, historical collection information, historical repayment information, historical credit clearing information, basic information and the like of a collection target, wherein each dimension comprises a certain number of variables. The raw data is the data collected from the different dimensions and not processed.
Further, the sub-module for processing and refining the collected raw data includes:
a unit for counting different parameter information of information event according to time interval and outputting information value;
a unit for counting different parameter information of information event according to different client states and outputting information value;
and the unit is used for aggregating the information values to different statistical granularities for numerical processing.
Those skilled in the art can understand that the parameter information includes the occurrence frequency, duration and amount of the above information events, and the numerical processing mode includes addition, maximum value calculation, minimum value calculation and the like; the processing and refining process refers to a process of carrying out secondary construction on the collected original data, and comprises the steps of counting the times, duration and amount of occurrence of the information events within a fixed time interval from the present, counting the times, duration and amount of occurrence of the information events according to the state of a client, aggregating the information values to different statistical granularities, and carrying out addition, maximum value calculation, minimum value calculation and the like.
Further, the module for processing the multidimensional information data of the collection object comprises:
the submodule is used for processing extreme values or abnormal values in the multi-dimensional information data;
the submodule is used for processing missing values in the multi-dimensional information data;
and the submodule is used for processing the category type variable data in the multi-dimensional information data and replacing each category in the category type variable with a numerical value.
Further, the sub-module for processing extreme values or abnormal values in the multi-dimensional information data includes:
the unit is used for determining an upper limit value and a lower limit value of the catalytic recovery target variable;
when each catalytic recovery target variable is valued, if the value is greater than the upper bound value, the value is re-assigned as the upper bound value; if the numerical value is smaller than the lower bound value, the numerical value is reassigned as the lower bound value; and the other cases are unchanged.
Further, the sub-module for processing missing values in the multi-dimensional information data includes:
a unit for filling missing values in each variable by an average value of the variable when processing numerical variable data;
and a unit for filling missing values in each variable by the mode of the variable when processing variable data of the type.
As can be understood by those skilled in the art, the embodiment utilizes an exception handler, a filler and an encoder in the system to perform data processing on the multidimensional information data of the collection object. The exception handler comprises a process for processing extreme values and abnormal values in the multi-dimensional information data, and the main steps are as follows: for a variable to be processed, determining an upper bound value and a lower bound value of the variable; and aiming at the value of each variable of the collection target, if the value is greater than the upper bound value, the variable is reassigned as the upper bound value, if the value is less than the lower bound value, the variable is reassigned as the lower bound value, and the other conditions are not changed. The stuffer processes missing values in the multi-dimensional information data, and the main steps are as follows: for numerical variable data, the missing value in each variable is filled with the average value of the variable; for categorical variable data, the missing value in each variable is filled with the mode of the variable. The encoder processes the categorical variable data in the multi-dimensional information data, and replaces each category in the categorical variable with a numerical value, such as "male" equal to 1 and "female" equal to 2.
And performing variable screening on the processed multi-dimensional information data by using a variable screener, and rejecting variables with low payment relevance between each dimension and the client and variables with strong correlation between each dimension and relatively low payment relevance between the same dimension and the client.
Further, the module for performing score prediction on the screened variables and dynamically urging to automatically judge the risk of the customer includes:
a submodule for dividing the data of the screened variables into reference samples and non-reference samples;
the step of judging whether the screened variables are reference samples or not, and if the variables are the reference samples, acquiring the customer data rules of the reference samples through an algorithm and predicting and scoring; if the sample is a non-reference sample, performing prediction scoring;
and the submodule is used for segmenting the clients according to the prediction scores and performing dynamic collection.
One skilled in the art will appreciate that the data for the screened variables described above are divided into reference samples and non-reference samples. The reference sample is the customer data with the collection record and payment label, and is used for training the model. The non-reference sample is data which has no payment label and needs to evaluate the payment willingness of the customer, namely the latest batch of data related to the customer who needs to implement the charging method. And if the sample is the reference sample, putting the reference sample into a learner to obtain a prediction scorer. The learner acquires data rules of good and bad clients in the reference sample through a machine learning algorithm, the data rules comprise data characteristics of the clients which are biased to repayment in all dimensions, data characteristics of the clients which are biased to overdue in all dimensions, frequency distribution of the good and bad clients which are calibrated as collection-promoting targets by the dynamic capturer and the like, and the data rules are formed into a prediction scoring device by the learner. If the sample is a non-reference sample, the non-reference sample is directly scored by the prediction scorer, a client with a high repayment willingness is biased to get a high score without passing through a learner, and a client which is easy to overdue is biased to get a low score. Aiming at the customers with different scores, different collection urging strategies are implemented, and for the customers with high section, namely the customers with strong repayment willingness, the strategies of short-term low-frequency intelligent voice call collection urging or low-frequency short message reminding and later-period low-frequency artificial telephone collection urging are adopted; for low-segment customers, namely customers who are easy to be overdue, the strategies of short-term high-frequency intelligent voice call collection or high-frequency short message reminding and later-term high-frequency manual call collection are adopted.
The collection hastening method provided by the embodiment is beneficial to discovering high-risk customers as early as possible and reducing bad account loss; unifying the collection strategy standard; the catalytic recovery efficiency is improved. The high-risk customers predicted by the collection-urging score are listed as key collection-urging targets, and collection-urging force is increased by adopting collection-urging modes such as increasing collection-urging frequency and getting home in advance, so that the customers can be promoted to pay as soon as possible. And (4) carrying out judicial collection or ex-commission collection on overdue clients without repayment willingness among high-risk clients as soon as possible. The customer with different risks is distinguished by the collection grading, and then the customer is assigned to different types of collectors by the system, so that the consistency, objectivity and unbiased property of the collection mode and the risk characteristics of the customer are ensured. Meanwhile, the risk preference is adjusted at any time according to the grading threshold value. When the economic cycle or credit cycle changes, the receiving acceleration can be increased in time by adjusting the system, and the unified regulation and control of the receiving acceleration strategy can be realized. The personal credit service is characterized by large service volume and small single amount. The collection prompting method adopts a mode of automatically sending short messages and letters or making voice calls to low-risk customers to prompt the customers to pay on the premise of controllable risk, and does not prompt the customers who can actively pay to collect in the early stage. The collection urging method realizes the automatic management of personal credit business, improves the collection urging efficiency and reduces the operation cost and the labor cost. Meanwhile, the customer satisfaction is correspondingly improved because the high-quality customers are not disturbed.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (14)
1. A dynamic collection method based on multi-dimensional information data is characterized by comprising the following steps:
s1 is used for dynamically capturing dynamic collection target;
s2 is used for acquiring multi-dimensional information data of the collection object;
s3 is used for carrying out data processing on the multidimensional information data of the collection object;
s4 is used for carrying out variable screening on the processed multidimensional information data;
and S5 is used for carrying out scoring prediction on the screened variables and carrying out dynamic collection on the automatically judged risk of the client.
2. The method for dynamically hastening receipts based on multidimensional information data as recited in claim 1, wherein the step of S2 for obtaining the multidimensional information data of the receiving target comprises: s21, acquiring original data of different-dimension collection targets; and S22 is used for processing and refining the collected original data.
3. The dynamic collection method based on multi-dimensional information data as claimed in claim 2, wherein the step of S22 for refining the collected raw data includes:
s221 is used for counting different parameter information of the information event according to the time interval and outputting the information value;
s222, counting different parameter information of the information event according to different client states and outputting an information value;
and S223, aggregating the information values to different statistical granularities for numerical processing.
4. The method for dynamically procuring multi-dimensional information data according to claim 3, wherein the step of S3 for processing the multi-dimensional information data of the procuring target includes:
s31 is used for processing extreme values or abnormal values in the multi-dimensional information data;
s32 is a step for processing missing values in the multidimensional information data;
s33 is a step of processing the categorical variable data in the multidimensional information data and replacing each category of the categorical variable with a numerical value.
5. The method for collecting data according to claim 4, wherein the step of S31 for processing extreme values or abnormal values in the multidimensional information data includes:
s311 is used for determining an upper limit value and a lower limit value of the catalytic recovery target variable;
s312, when the value of each catalytic target variable is taken, if the value is larger than the upper bound value, the value is re-assigned as the upper bound value; if the numerical value is smaller than the lower bound value, the numerical value is reassigned as the lower bound value; otherwise, the steps are unchanged.
6. The method for dynamically hastening harvesting based on multi-dimensional information data as claimed in claim 5, wherein the step of S32 for processing missing values in the multi-dimensional information data comprises:
s321, when processing numerical variable data, filling missing values in each variable by an average value of the variable;
s322 is a step of filling missing values in each variable with the mode of the variable when processing the variable data of the type.
7. The method for dynamically collecting data based on multidimensional information as recited in claim 6, wherein the step S5 for performing score prediction on the filtered variables and performing dynamic collection on the automatically judged risk of the customer comprises:
s51 is used for dividing the data of the screened variable into a reference sample and a non-reference sample;
s52 is used for judging whether the screened variables are reference samples or not, and if the screened variables are the reference samples, the customer data rules of the reference samples are obtained through an algorithm and the prediction and scoring steps are carried out; if the sample is a non-reference sample, performing prediction and scoring;
and S53 is used for segmenting and dividing the client according to the prediction score and performing dynamic collection.
8. A dynamic collection system based on multi-dimensional information data is characterized by comprising:
a module for dynamically capturing dynamic collection targets;
the module is used for acquiring multi-dimensional information data of the collection target;
the module is used for carrying out data processing on the multidimensional information data of the collection object;
the module is used for carrying out variable screening on the processed multidimensional information data;
and the module is used for grading and predicting the screened variables and carrying out dynamic collection on different clients.
9. The multidimensional information data based dynamic collection system of claim 8, wherein the module for obtaining multidimensional information data of collection targets comprises:
a submodule for collecting original data of different-dimension collection targets;
and the submodule is used for processing and refining the collected original data.
10. The multidimensional information data based dynamic collection system of claim 9, wherein the sub-module for refining the collected raw data comprises:
a unit for counting different parameter information of information event according to time interval and outputting information value;
a unit for counting different parameter information of information event according to different client states and outputting information value;
and the unit is used for aggregating the information values to different statistical granularities for numerical processing.
11. The multidimensional information data based dynamic collection system of claim 10, wherein the means for data processing the multidimensional information data of the collection target comprises:
the submodule is used for processing extreme values or abnormal values in the multi-dimensional information data;
the submodule is used for processing missing values in the multi-dimensional information data;
and the submodule is used for processing the category type variable data in the multi-dimensional information data and replacing each category in the category type variable with a numerical value.
12. The system for multi-dimensional information data-based dynamic collection according to claim 11, wherein the sub-module for processing extreme values or abnormal values in the multi-dimensional information data comprises:
the unit is used for determining an upper limit value and a lower limit value of the catalytic recovery target variable;
when each catalytic recovery target variable is valued, if the value is greater than the upper bound value, the value is re-assigned as the upper bound value; if the numerical value is smaller than the lower bound value, the numerical value is reassigned as the lower bound value; and the other cases are unchanged.
13. The multidimensional information data based dynamic harvesting system of claim 12, wherein the sub-module for processing missing values in multidimensional information data comprises:
a unit for filling missing values in each variable by an average value of the variable when processing numerical variable data;
and a unit for filling missing values in each variable by the mode of the variable when processing variable data of the type.
14. The system of claim 13, wherein the means for dynamically collecting and forecasting scores for the filtered variables and automatically collecting the automated assessment of customer risk comprises:
a submodule for dividing the data of the screened variables into reference samples and non-reference samples;
the step of judging whether the screened variables are reference samples or not, and if the variables are the reference samples, acquiring the customer data rules of the reference samples through an algorithm and predicting and scoring; if the sample is a non-reference sample, performing prediction scoring;
and the submodule is used for segmenting the clients according to the prediction scores and performing dynamic collection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010369251.2A CN111581252A (en) | 2020-05-04 | 2020-05-04 | Dynamic collection urging method and system based on multi-dimensional information data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010369251.2A CN111581252A (en) | 2020-05-04 | 2020-05-04 | Dynamic collection urging method and system based on multi-dimensional information data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111581252A true CN111581252A (en) | 2020-08-25 |
Family
ID=72126239
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010369251.2A Pending CN111581252A (en) | 2020-05-04 | 2020-05-04 | Dynamic collection urging method and system based on multi-dimensional information data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111581252A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065945A (en) * | 2021-03-17 | 2021-07-02 | 上海浦东发展银行股份有限公司 | Method and system for classifying repayment willingness of customer for collection, verification and sale |
| WO2024059019A1 (en) * | 2022-09-14 | 2024-03-21 | Nec Laboratories America, Inc. | Multi-modality data augmentation engine to improve rare driving scenario detection for vehicle sensors |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106952155A (en) * | 2017-03-08 | 2017-07-14 | 深圳前海纵腾金融科技服务有限公司 | A kind of collection method and device based on credit scoring |
| CN109509086A (en) * | 2018-11-28 | 2019-03-22 | 上海点融信息科技有限责任公司 | The method, apparatus and storage medium of processing collection business based on artificial intelligence |
| CN109658043A (en) * | 2018-10-16 | 2019-04-19 | 深圳壹账通智能科技有限公司 | Collection data processing method, device, equipment and storage medium |
| CN110288460A (en) * | 2019-04-24 | 2019-09-27 | 武汉众邦银行股份有限公司 | Collection prediction technique, device, equipment and storage medium based on propagated forward |
| CN110909984A (en) * | 2019-10-28 | 2020-03-24 | 苏宁金融科技(南京)有限公司 | Business data processing model training method, business data processing method and device |
-
2020
- 2020-05-04 CN CN202010369251.2A patent/CN111581252A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106952155A (en) * | 2017-03-08 | 2017-07-14 | 深圳前海纵腾金融科技服务有限公司 | A kind of collection method and device based on credit scoring |
| CN109658043A (en) * | 2018-10-16 | 2019-04-19 | 深圳壹账通智能科技有限公司 | Collection data processing method, device, equipment and storage medium |
| CN109509086A (en) * | 2018-11-28 | 2019-03-22 | 上海点融信息科技有限责任公司 | The method, apparatus and storage medium of processing collection business based on artificial intelligence |
| CN110288460A (en) * | 2019-04-24 | 2019-09-27 | 武汉众邦银行股份有限公司 | Collection prediction technique, device, equipment and storage medium based on propagated forward |
| CN110909984A (en) * | 2019-10-28 | 2020-03-24 | 苏宁金融科技(南京)有限公司 | Business data processing model training method, business data processing method and device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065945A (en) * | 2021-03-17 | 2021-07-02 | 上海浦东发展银行股份有限公司 | Method and system for classifying repayment willingness of customer for collection, verification and sale |
| WO2024059019A1 (en) * | 2022-09-14 | 2024-03-21 | Nec Laboratories America, Inc. | Multi-modality data augmentation engine to improve rare driving scenario detection for vehicle sensors |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106952155A (en) | A kind of collection method and device based on credit scoring | |
| CN112766550B (en) | Random forest-based power failure sensitive user prediction method, system, storage medium and computer equipment | |
| CN109191283A (en) | Method for prewarning risk and system | |
| CN111324862A (en) | Method and system for monitoring behavior in loan | |
| CN110837963A (en) | Risk control platform construction method based on data, model and strategy | |
| CN103024762A (en) | Service feature based communication service forecasting method | |
| CN111581252A (en) | Dynamic collection urging method and system based on multi-dimensional information data | |
| CN110909984A (en) | Business data processing model training method, business data processing method and device | |
| CN111275338A (en) | Method, device, equipment and storage medium for judging enterprise fraud behaviors | |
| CN112418738A (en) | Staff operation risk prediction method based on logistic regression | |
| CN110413901B (en) | Credit risk assessment method based on social network analysis | |
| CN116596663A (en) | Method and device for establishing business client credit scoring model | |
| CN111061948A (en) | User label recommendation method and device, computer equipment and storage medium | |
| CN112686446A (en) | Machine learning interpretability-oriented credit default prediction method and system | |
| CN117011031A (en) | Intelligent financial risk monitoring and early warning platform using big data information processing technology | |
| CN117371727A (en) | Intelligent agricultural monitoring and planning method and system based on Internet of things | |
| CN111738824A (en) | Method, device and system for screening financial data processing modes | |
| CN112613971A (en) | User credit risk assessment method and device, computer equipment and storage medium | |
| CN116843446A (en) | Method for evaluating and attributing infraction risk of small amount credit based on interpretable machine learning | |
| CN111951105A (en) | Intelligent credit wind control system based on multidimensional big data analysis | |
| CN110738565A (en) | Real estate finance artificial intelligence composite wind control model based on data set | |
| CN116091106A (en) | Multifunctional financial cost evaluation system | |
| CN113592140A (en) | Electric charge payment prediction model training system and electric charge payment prediction model | |
| CN114090385A (en) | Monitoring and early warning method, device and equipment for service running state | |
| CN113421155A (en) | Steel trade enterprise portrait multidimensional model construction method, wind control management method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200825 |
|
| RJ01 | Rejection of invention patent application after publication |