CN-122022238-A - Dynamic multi-rule intelligent case dividing platform for poor case of individual credit

CN122022238ACN 122022238 ACN122022238 ACN 122022238ACN-122022238-A

Abstract

The invention discloses a dynamic multi-rule intelligent case division platform for bad asset cases, which belongs to the field of financial bad asset management and data processing and comprises an initialization module, a dynamic rule engine module, a batch verification module, a weight calculation module, an allocation decision module, an execution module and a balancing module, wherein the platform firstly acquires three-dimensional upper limit and load data, calculates a global pressure load index and dynamically determines a rule priority sequence. After distribution, the platform calculates a multidimensional fairness index, and if the multidimensional fairness index is incorrect, the platform triggers a directional fine adjustment operation. The invention adopts dynamic priority adjustment, predictive weight calculation and closed-loop self-learning gradient mechanism, can sense the load of the asset handling executor in real time, accurately constrains complex business, and improves the fairness of case distribution, the self-adaptive capacity of the system and the overall operation efficiency.

Inventors

JIANG JIAWEN
GU GANG
HE PENGFEI

Assignees

广东横琴深水云科数字科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251217

Claims (10)

1. The dynamic multi-rule intelligent case division platform for the poor case of each credit is characterized by comprising an initialization module, a dynamic rule engine module, a batch verification module, a weight calculation module, an allocation decision module, an execution module and a balancing module; The initialization module is used for acquiring the upper limit of the three-dimensional limit of the asset handling executor, the rule influence coefficient, the current load data and the case pool to be distributed, and generating initialization configuration data, wherein the three-dimensional limit is the number, the amount and the number of houses; The dynamic rule engine module is used for calculating a global load pressure index based on the current load data in the initialization configuration data, dynamically determining a rule priority sequence according to the global load pressure index and generating a dynamic rule sequence; the batch verification module is used for executing three-dimensional limit verification according to the dynamic rule sequence aiming at the cases to be distributed to generate a candidate asset handling executor set; The weight calculation module is used for calculating the residual resource duty ratio of each asset handling executor in the candidate asset handling executor set by combining the current load data and the upper limit of the three-dimensional limit, calculating the distribution weight by combining the current global load pressure index change prediction based on the rule influence coefficient, and generating an asset handling executor weight list; The allocation decision module is used for generating a winning candidate set according to the asset handling executor weight list and a scaling factor preset based on allocation weights, randomly selecting a target asset handling executor from the winning candidate set and generating a case allocation instruction; The execution module is used for executing the case allocation instruction, allocating the case to a target asset handling executor and updating load data of the asset handling executor; and the balancing module is used for calculating multidimensional fairness indexes based on the updated asset handling executor load data, generating fairness evaluation results and triggering directional fine adjustment operation according to the fairness evaluation results.
2. The dynamic multi-rule intelligent case-dividing platform for poor-credit cases according to claim 1, wherein the initialization module comprises an upper limit acquisition unit, an influence factor configuration unit and a case data acquisition unit; an upper limit acquisition unit configured to acquire an upper limit of the number, amount, and number of users of a prescribed asset handling executor; the influence coefficient acquisition unit is used for acquiring the number dimension influence coefficient, the sum dimension influence coefficient and the number of units dimension influence coefficient in the rule influence coefficient; The case data acquisition unit is used for acquiring case attribute information in the case pool to be distributed, wherein the attribute information comprises a client unique identifier and case amount.
3. The dynamic multi-rule intelligent case-oriented platform for poor case credits of claim 1, wherein the dynamic rule engine module comprises a load data acquisition unit, an asset handling executor load variance calculation unit and a pressure index calculation unit; The load data acquisition unit is used for acquiring an asset handling execution case number set, an asset handling execution case total amount set and an asset handling execution person bearing independent client number set in the current load data; The asset handling executor load variance calculation unit is used for calculating variances of the appointed asset handling executor case number set, the asset handling executor case total sum set and the asset handling executor load independent client number set respectively to generate case number dimension load variances, sum dimension load variances and independent client number dimension load variances; And the pressure index calculation unit is used for carrying out weighted summation after normalizing the case number dimension load variance, the sum dimension load variance and the independent customer number dimension load variance to generate the global load pressure index.
4. The dynamic multi-rule intelligent case-division platform for poor case-on-credit according to claim 1, wherein the batch verification module comprises a case and asset handling executor information acquisition unit, a constraint verification unit and a candidate set generation unit; The case and asset handling executor information acquisition unit is used for acquiring the unique client identification, the case amount and the case number of the cases to be allocated, and the current data load and the three-dimensional limit upper limit of the candidate asset handling executor; The constraint verification unit is used for verifying whether the number of the households, the amount of money and the number of the cases to be allocated are all not more than the upper limit of the three-dimensional limit after the cases to be allocated are allocated to the candidate asset handling executives according to the priority sequence of the dynamic rule sequence; and the candidate set generating unit is used for bringing the candidate asset handling executives into the candidate asset handling executives only when the number of the households, the sum of money and the number of the households do not exceed the upper limit of the three-dimensional limit.
5. The dynamic multi-rule intelligent case-division platform for poor case of credit according to claim 2, wherein the weight calculation module comprises an initial weight calculation unit and a predicted value unit; the initial weight calculation unit is used for generating an initial weight value based on the quantity dimension influence coefficient, the sum dimension influence coefficient and the household number dimension influence coefficient and the residual resource duty ratio; And the predicted value unit is used for subtracting the product of the current global load pressure index change prediction and the adjustment coefficient from the initial weight value to generate an allocation weight, wherein the adjustment coefficient is used for controlling the influence intensity of the global load pressure index change prediction on the final allocation weight.
6. The dynamic multi-rule intelligent case-oriented division platform for poor case of credit according to claim 1, wherein the allocation decision module comprises a dynamic selection threshold generating unit and a winning candidate set screening unit; A dynamic selection threshold generating unit, configured to obtain an optimized weight value in the asset handling executor weight list, and generate a dynamic selection threshold in combination with a scaling factor preset based on the allocation weight; and the winning candidate set screening unit is used for selecting asset handling executives with the distribution weights greater than or equal to the dynamic selection threshold value from the asset handling executives weight list to generate winning candidate sets.
7. The dynamic multi-rule intelligent case-oriented division platform for poor case of credit according to claim 1, wherein the balancing module comprises an unbalance index identification unit, an excess screening unit and a power consumption distribution unit; The unbalanced index identification unit is used for identifying dominant unbalanced dimensions based on the multidimensional fairness index and generating a dominant dimension identification result when the fairness evaluation result does not meet a preset multidimensional fairness index qualification threshold; The excess screening unit is used for screening excess asset handling executor sets according to the dominant dimension identification result, screening movable cases from the excess asset handling executor sets and generating movable case sets; And the power consumption distribution unit is used for distributing the cases in the movable case set to the receiving asset handling executor based on the updated asset handling executor load data, and generating a fine adjustment distribution result.
8. The dynamic multi-rule intelligent case-splitting platform for poor-credit cases according to claim 1, wherein the unbalance index identifying unit comprises: acquiring a quantity dimension balance rate, an amount dimension matching degree and a household number dimension concentration degree in the multidimensional fairness index; And comparing the number dimension balance rate, the sum dimension matching degree and the number of units dimension concentration degree, and generating the dimension with the lowest value as the dominant unbalanced dimension.
9. The dynamic multi-rule intelligent case-division platform for poor case of credit according to claim 1, wherein the platform further comprises a data history recording unit and a parameter self-updating unit; the data history recording unit is used for recording the fine-tuned root cause data, fine-tuned action data and effect evaluation data and generating history data; and the parameter self-updating unit is used for updating the calculation parameters and the adjustment coefficients of the global load pressure index by using the historical data to generate updated parameters.
10. A dynamic multi-rule intelligent case division method for poor case of credit, characterized in that the method applies a dynamic multi-rule intelligent case division platform for poor case of credit according to any one of claims 1-9, the method comprises: acquiring the upper limit of the three-dimensional limit of an asset handling executor, a rule influence coefficient, current load data and a case pool to be allocated, and generating initialization configuration data, wherein the three-dimensional limit is the number, the amount and the number of houses; Calculating a global load pressure index based on current load data in the initialization configuration data, dynamically determining a rule priority sequence according to the global load pressure index, and generating a dynamic rule sequence; Executing three-dimensional limit verification according to the dynamic rule sequence aiming at the case to be distributed to generate a candidate asset handling executor set; Calculating the residual resource duty ratio of each asset handling executor in the candidate asset handling executor set by combining the current load data and the upper limit of the three-dimensional limit, calculating distribution weights by combining the current global load pressure index change prediction based on the rule influence coefficient, and generating an asset handling executor weight list; Generating a winning candidate set according to the asset handling executor weight list and a scale factor preset based on allocation weights, randomly selecting a target asset handling executor from the winning candidate set, and generating a case allocation instruction; Executing the case allocation instruction, allocating the case to a target asset handling executor, and updating asset handling executor load data; and calculating a multidimensional fairness index based on the updated asset handling executor load data, generating a fairness evaluation result, and triggering a directional fine adjustment operation according to the fairness evaluation result.

Description

Dynamic multi-rule intelligent case dividing platform for poor case of individual credit Technical Field The invention relates to the field of financial bad asset management and data processing, in particular to a dynamic multi-rule intelligent case division platform for bad cases. Background The collection management of individual bad cases is a key link of financial institution risk control and asset security. In the process, mass cases to be collected are reasonably and efficiently distributed to asset handling executives as starting points of the whole collection flow, and the advantages and disadvantages of the distribution strategy directly influence collection efficiency, recovery rate and utilization rate of human resources. Therefore, various computer-aided intelligent scheme dividing platforms are developed in the industry, and the aim is to replace the traditional manual scheme dividing by an automatic system so as to meet the large-scale and complicated business demands. In the prior art, case distribution systems typically employ distribution logic based on fixed rules. Common schemes include poll allocation, i.e., sequentially assigning cases to each asset handling executor in a preset order, or employing a single capacity-based allocation scheme, e.g., assigning new cases to the least number of asset handling executives based solely on the number of cases currently held by the asset handling executor. Some somewhat complex systems may introduce simple matching rules for attributes such as case amounts or territories, but these rules, once set, typically remain unchanged for longer periods. However, the prior art solutions described above have significant limitations. Firstly, static allocation rules lack the perception capability of dynamic change of the overall load of a collection team, cannot adapt to business scenes of real-time fluctuation of case inflow and type structures, and easily cause load unbalance in certain dimensions. Secondly, the distribution decision is made only by relying on a single dimension, such as the number of cases, neglecting important load dimensions such as case amount, customer distribution, etc., which may cause a large difference in actual working pressure and risk exposure between asset handling executives. In addition, traditional allocation decisions tend to be local and immediate, lack consideration of the long-term impact of single allocation behavior on the global balance of the system, and tend to accumulate and amplify imbalance conditions during long-term operation. Disclosure of Invention In order to solve the problems, the invention provides a dynamic multi-rule intelligent case division platform for poor credit cases, which adopts the technical means of dynamically adjusting rule priority based on global load pressure index, calculating allocation weight by combining global influence prediction and performing closed-loop fine adjustment based on fairness indexes after allocation, so that dynamic and prospective balanced allocation of cases in multiple dimensions of quantity, amount and household number can be realized, and the intelligence level and the overall operation fairness of case division are improved. In order to achieve the above purpose, the application adopts the following technical scheme: in the first aspect, a dynamic multi-rule intelligent case division platform for poor cases of credit is provided, which comprises an initialization module, a dynamic rule engine module, a batch verification module, a weight calculation module, an allocation decision module, an execution module and a balancing module, wherein: The initialization module is used for acquiring the upper limit of the three-dimensional limit of the asset handling executor, the rule influence coefficient, the current load data and the case pool to be distributed, and generating initialization configuration data, wherein the three-dimensional limit is the number, the amount and the number of houses; The dynamic rule engine module is used for calculating a global load pressure index based on the current load data in the initialization configuration data, dynamically determining a rule priority sequence according to the global load pressure index and generating a dynamic rule sequence; the batch verification module is used for executing three-dimensional limit verification according to the dynamic rule sequence aiming at the cases to be distributed to generate a candidate asset handling executor set; The weight calculation module is used for calculating the residual resource duty ratio of each asset handling executor in the candidate asset handling executor set by combining the current load data and the upper limit of the three-dimensional limit, calculating the distribution weight by combining the current global load pressure index change prediction based on the rule influence coefficient, and generating an asset handling executor weight list; The allocation decision module