Negotiation state-based non-performing asset disposal scheme generation method and system

Abstract

This invention discloses a method and system for generating non-performing asset disposal plans based on negotiation states. The method includes: acquiring original events from multiple rounds of negotiation between the user and the disposal entity; encoding the original events into event units and constructing a negotiation event sequence in chronological order; constructing a negotiation state diagram based on the negotiation event sequence; determining the user's current negotiation state based on the evolution of the negotiation state diagram; generating candidate parameter combinations for the disposal plan in a preset parameter search space based on the boundary constraints of the negotiation state; outputting the acceptance probability and performance probability of each candidate parameter combination based on a preset prediction network; and determining the target parameter combination of the disposal plan based on the acceptance probability and performance probability. Therefore, this invention effectively utilizes feedback from multiple rounds of negotiation, significantly improving the timeliness and universality of generated disposal plans.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for generating non-performing asset disposal plans based on negotiation status. Background Technology

[0002] In the context of negotiated disposal of non-performing assets, disposal plans are often a crucial means to facilitate transactions and recovery. Existing technologies primarily rely on the experience of sales representatives, fixed policy tables, or aging-based rules to offer debtors various relief options, such as lump-sum settlement, interest reduction, penalty reduction, down payment with installments, and installments after extension. Among these, effectively utilizing feedback from multiple rounds of negotiation becomes a key issue in generating disposal plans.

[0003] Existing technologies generally suffer from the following drawbacks: (1) The scheme generation is mainly based on single-point-of-time data, ignoring the feedback evolution of the debtor in multiple rounds of negotiation; (2) It is difficult to uniformly model the debtor's behavioral characteristics (such as price-cutting behavior, delaying behavior, condition exchange behavior, emotional changes, and the credibility of the promise); (3) Usually only the negotiation rate or short-term payment is focused on, and there is still a lack of joint optimization for the stability of subsequent performance; (4) Different agents often adopt different strategies for similar cases, making it difficult to accumulate reusable negotiation experience; (5) The lack of a standardized mechanism for advancing, maintaining, switching and terminating the solution in terms of time results in low efficiency in actual solution generation and affects customer experience.

[0004] It is evident that existing technologies have shortcomings that urgently need to be addressed. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for generating non-performing asset disposal plans based on negotiation status, which can effectively utilize multi-round negotiation feedback and significantly improve the timeliness and universality of generating disposal plans.

[0006] To address the aforementioned technical problems, the first aspect of this invention discloses a method for generating non-performing asset disposal plans based on negotiation status, the method comprising: The original events between the user and the handling entity during multiple rounds of negotiation are obtained, and the original events are encoded into event units to construct a negotiation event sequence in chronological order. A negotiation state diagram is constructed based on the negotiation event sequence, and the user's current negotiation state is determined based on the evolution of the negotiation state diagram. Based on the boundary constraints of the negotiation state, candidate parameter combinations for the disposal plan are generated in the preset parameter search space. Based on the preset prediction network, the acceptance probability and fulfillment probability of each candidate parameter combination are output, and the target parameter combination of the disposal plan is determined according to the acceptance probability and fulfillment probability.

[0007] As an optional implementation, in the first aspect of the present invention, the original event includes one or more of the following: outbound call event, SMS arrival event, online chat event, self-negotiated page operation event, user callback event, commitment registration event, commitment default event, historical performance event, and human agent remarks event; Negotiation status includes probing status, price reduction status, condition exchange status, pending transaction status, delay status, and low credibility status; The parameter combinations include plan type, reduction / exemption ratio, down payment ratio, number of installments, amount per installment, and plan validity period.

[0008] As an optional implementation, in the first aspect of the present invention, the original event encoding obtains event information from the original event through semantic recognition, entity extraction and numerical extraction, and maps the event information into structured event units; The event information includes the event timestamp, contact channel type, reporting value, counter-reporting value, response delay, reason for refusal, promised amount, promised date, sentiment, and performance result.

[0009] As an optional implementation, in a first aspect of the invention, the negotiation state diagram includes stage nodes and state transition edges; The stage node is the negotiation stage corresponding to the event unit in the negotiation event sequence; The state transition edge represents the state transition relationship between nodes in each stage during the negotiation stage. The state transition relationship is the driving relationship between one or more of the driving factors such as changes in price difference, changes in response delay, changes in sentiment, or changes in the credibility of commitment and the state changes during the negotiation stage. The changes in the price difference, response delay, sentiment, or credibility of the commitment are determined by the corresponding slope changes calculated in the event information.

[0010] As an optional implementation, in the first aspect of the invention, determining the user's current negotiation state based on the evolution of the negotiation state diagram includes: A sliding window is constructed by negotiating event units in a sequence of events. The event units in the negotiation state diagram are statistically analyzed using a sliding window to calculate the mean and slope of change of the driving factors in the state transition relationship, and state identification features are constructed based on the mean and slope of change. The candidate state score is obtained by weighting and summing the state recognition features with state weights and bias terms according to the preset state scoring function. Calculate the candidate state score difference between adjacent sliding windows, and construct a state transition matrix by combining the candidate state score difference and the number of transitions. The scores of each candidate state are smoothed by using the state transition matrix, and the candidate state with the highest score among the candidate states that meet the smoothing criteria is determined as the user's current negotiation state.

[0011] As an optional implementation, in the first aspect of the invention, based on the boundary constraints of the negotiated state, candidate parameter combinations for the disposal scheme are generated in a preset parameter search space, including: The basic parameters for determining the parameter search space are obtained from regulatory policies, business strategies, and asset categories. The parameter search space is then constructed based on the actual parameters from historically handled cases. Based on the negotiation status, determine the corresponding scheme type of the disposal plan, and based on the scheme type, determine the parameter boundary of the candidate parameters as the parameter search space, and output the state boundary constraints of the feasible solution corresponding to the candidate parameter combination. Based on the state boundary constraints, feasible solutions for the corresponding parameter combinations of the treatment scheme are obtained by filtering in the parameter search space, forming multiple candidate parameter combinations.

[0012] As an optional implementation, in the first aspect of the invention, the acceptance probability and fulfillment probability of each candidate parameter combination are output based on a preset prediction network, including: The specific steps for outputting the acceptance probability are as follows: Extract the difference feature between the report value of the candidate parameter combination and the counter-report value proposed by the user in the previous period; Extract matching features between candidate parameter combinations and users' historical affordable amounts; Extract the acceptance bias feature corresponding to the current negotiation state; The gap feature, matching feature, and acceptance bias feature are used as inputs to the acceptance probability prediction network, and the output is the user's acceptance probability for each candidate parameter combination. The acceptance probability prediction network is trained using a training dataset that includes multiple candidate parameter combinations and original events, as well as a historical case dataset with acceptance probability labels. The specific steps for outputting the fulfillment probability are as follows: The candidate parameter combination and the user's historical performance record are used as input to the performance probability prediction network. The output is the performance probability of the user for each candidate parameter combination. The performance probability prediction network is trained by a training dataset including multiple candidate parameter combinations and original events, as well as a historical case dataset with performance probability labels. The acceptance probability prediction network and the performance probability prediction network were trained together using a public training dataset with different prediction result labels.

[0013] As an optional implementation, in the first aspect of the invention, determining the target parameter combination of the disposal plan based on the acceptance probability and the fulfillment probability includes: The candidate parameter combinations of the disposal plan are sorted from largest to smallest according to the probability of acceptance and the probability of fulfillment; The acceptance probabilities of the candidate parameter combinations with the highest probability of performance are compared in turn. The candidate parameter combination with the highest probability of acceptance under the condition of the highest probability of performance is selected to obtain the target parameter combination of the disposal plan. Candidate parameter combinations whose probability of performance is no greater than that of the target parameter combination and whose probability of acceptance is no greater than that of the target parameter combination are considered as alternative solutions to the target parameter combination.

[0014] A second aspect of this invention discloses a system for generating non-performing asset disposal plans based on negotiation status, the system comprising: The acquisition module is used to acquire the original events between the user and the handling entity in multiple rounds of negotiation, encode the original events into event units, and construct a negotiation event sequence according to the time sequence. The state module is used to construct a negotiation state diagram based on the negotiation event sequence, and to determine the user's current negotiation state based on the evolution of the negotiation state diagram. The generation module is used to generate candidate parameter combinations for disposal schemes in a preset parameter search space based on the boundary constraints of the negotiation state. The determination module is used to output the acceptance probability and fulfillment probability of each candidate parameter combination based on the preset prediction network, and to determine the target parameter combination of the disposal plan based on the acceptance probability and fulfillment probability.

[0015] As an optional implementation, in the second aspect of the present invention, the original event includes one or more of the following: outbound call event, SMS arrival event, online chat event, self-negotiated page operation event, user callback event, commitment registration event, commitment default event, historical performance event, and human agent remarks event; Negotiation status includes probing status, price reduction status, condition exchange status, pending transaction status, delay status, and low credibility status; The parameter combinations include plan type, reduction / exemption ratio, down payment ratio, number of installments, amount per installment, and plan validity period.

[0016] As an optional implementation, in a second aspect of the invention, the original event encoding obtains event information from the original event through semantic recognition, entity extraction and numerical extraction, and maps the event information into structured event units; The event information includes the event timestamp, contact channel type, reporting value, counter-reporting value, response delay, reason for refusal, promised amount, promised date, sentiment, and performance result.

[0017] As an optional implementation, in a second aspect of the invention, the negotiation state diagram includes stage nodes and state transition edges; The stage node is the negotiation stage corresponding to the event unit in the negotiation event sequence; The state transition edge represents the state transition relationship between nodes in each stage during the negotiation stage. The state transition relationship is the driving relationship between one or more of the driving factors such as changes in price difference, changes in response delay, changes in sentiment, or changes in the credibility of commitment and the state changes during the negotiation stage. The changes in the price difference, response delay, sentiment, or credibility of the commitment are determined by the corresponding slope changes calculated in the event information.

[0018] As an optional implementation, in a second aspect of the invention, determining the user's current negotiation state based on the evolution of the negotiation state diagram includes: A sliding window is constructed by negotiating event units in a sequence of events. The event units in the negotiation state diagram are statistically analyzed using a sliding window to calculate the mean and slope of change of the driving factors in the state transition relationship, and state identification features are constructed based on the mean and slope of change. The candidate state score is obtained by weighting and summing the state recognition features with state weights and bias terms according to the preset state scoring function. Calculate the candidate state score difference between adjacent sliding windows, and construct a state transition matrix by combining the candidate state score difference and the number of transitions. The scores of each candidate state are smoothed by using the state transition matrix, and the candidate state with the highest score among the candidate states that meet the smoothing criteria is determined as the user's current negotiation state.

[0019] As an optional implementation, in a second aspect of the invention, based on the boundary constraints of the negotiated state, candidate parameter combinations for the disposal scheme are generated in a preset parameter search space, including: The basic parameters for determining the parameter search space are obtained from regulatory policies, business strategies, and asset categories. The parameter search space is then constructed based on the actual parameters from historically handled cases. Based on the negotiation status, determine the corresponding scheme type of the disposal plan, and based on the scheme type, determine the parameter boundary of the candidate parameters as the parameter search space, and output the state boundary constraints of the feasible solution corresponding to the candidate parameter combination. Based on the state boundary constraints, feasible solutions for the corresponding parameter combinations of the treatment scheme are obtained by filtering in the parameter search space, forming multiple candidate parameter combinations.

[0020] As an optional implementation, in a second aspect of the invention, the acceptance probability and fulfillment probability of each candidate parameter combination are output based on a preset prediction network, including: The specific steps for outputting the acceptance probability are as follows: Extract the difference feature between the report value of the candidate parameter combination and the counter-report value proposed by the user in the previous period; Extract matching features between candidate parameter combinations and users' historical affordable amounts; Extract the acceptance bias feature corresponding to the current negotiation state; The gap feature, matching feature, and acceptance bias feature are used as inputs to the acceptance probability prediction network, and the output is the user's acceptance probability for each candidate parameter combination. The acceptance probability prediction network is trained using a training dataset that includes multiple candidate parameter combinations and original events, as well as a historical case dataset with acceptance probability labels. The specific steps for outputting the fulfillment probability are as follows: The candidate parameter combination and the user's historical performance record are used as input to the performance probability prediction network. The output is the performance probability of the user for each candidate parameter combination. The performance probability prediction network is trained by a training dataset including multiple candidate parameter combinations and original events, as well as a historical case dataset with performance probability labels. The acceptance probability prediction network and the performance probability prediction network were trained together using a public training dataset with different prediction result labels.

[0021] As an optional implementation, in a second aspect of the invention, determining the target parameter combination of the disposal plan based on the acceptance probability and the fulfillment probability includes: The candidate parameter combinations of the disposal plan are sorted from largest to smallest according to the probability of acceptance and the probability of fulfillment; The acceptance probabilities of the candidate parameter combinations with the highest probability of performance are compared in turn. The candidate parameter combination with the highest probability of acceptance under the condition of the highest probability of performance is selected to obtain the target parameter combination of the disposal plan. Candidate parameter combinations whose probability of performance is no greater than that of the target parameter combination and whose probability of acceptance is no greater than that of the target parameter combination are considered as alternative solutions to the target parameter combination.

[0022] A third aspect of this invention discloses another system for generating non-performing asset disposal plans based on negotiation status, the system comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the method for generating a non-performing asset disposal plan based on negotiation status disclosed in the first aspect of the present invention.

[0023] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the method for generating a non-performing asset disposal plan based on a negotiation state disclosed in the first aspect of the present invention.

[0024] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: This invention encodes the original events of multiple rounds of negotiation into event units and constructs a negotiation event sequence according to time sequence. This enables decision-making based on the dynamic evolution trajectory of the user throughout the entire process, effectively capturing the real changing trends of user intentions and capabilities, and avoiding misjudgments or missed opportunities due to ignoring intermediate feedback evolution. By abstracting the negotiation event sequence into a negotiation state graph, using nodes and edges in the graph structure to represent different states and their transition relationships in the negotiation process, the implicit user characteristics scattered in each round of negotiation are uniformly mapped into calculable and evolvable negotiation states. This overcomes the limitations of different agents judging user intentions based on subjective experience, and the inability to uniformly model strategies for similar cases. This provides a basis for subsequent solution generation. It provides standardized state inputs and also serves as a knowledge carrier for negotiation experience across agents and cases. By defining parameter boundaries based on the current negotiation state when generating a resolution plan, multiple candidate plans are generated within this constraint space. Then, a prediction network is used to estimate the probability of each candidate plan being accepted by the debtor and the probability of the debtor fulfilling its obligations on schedule. This achieves joint optimization of the dual objectives of transaction probability and performance stability, avoiding excessive concessions to facilitate a current transaction that could lead to a high default rate later on, and significantly improving the rationality and controllability of the resolution plan. The evolutionary logic of the negotiation state diagram enables the effective use of multi-round negotiation feedback, significantly improving the timeliness and universality of the generated resolution plan. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating a method for generating non-performing asset disposal plans based on negotiation status, as disclosed in an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of the structure of a non-performing asset disposal plan generation system based on negotiation status disclosed in an embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram of another non-performing asset disposal scheme generation system based on negotiation status disclosed in an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0031] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0032] This invention discloses a method and system for generating non-performing asset disposal plans based on negotiation states. By encoding the original events of multiple rounds of negotiation into event units and constructing a negotiation event sequence according to time sequence, the disposal plan can be decided based on the dynamic evolution trajectory of the user throughout the entire process. This effectively captures the real changing trends of user willingness and ability, avoiding misjudgment or missed window periods due to ignoring intermediate feedback evolution. By abstracting the negotiation event sequence into a negotiation state graph, the nodes and edges in the graph structure represent different states and their transition relationships in the negotiation process. This unifies the implicit user characteristics scattered in each round of negotiation into a computable and evolvable negotiation state, overcoming the limitations of different agents judging user intentions based on subjective experience, the differences in strategies for similar cases, and the inability to determine user intentions. Overcoming the limitations of unified modeling, this approach provides standardized state inputs for subsequent solution generation and serves as a knowledge carrier for negotiation experience across agents and cases. By defining parameter boundaries based on the current negotiation state when generating a solution, multiple candidate solutions are generated within this constrained space. A prediction network is then used to estimate the probability of each candidate solution being accepted by the debtor and the probability of the debtor fulfilling their obligations on schedule. This achieves joint optimization of the dual objectives of transaction probability and performance stability, avoiding excessive concessions to facilitate a current transaction that could lead to high default rates later, significantly improving the rationality and controllability of the solution. Furthermore, the evolutionary logic of the negotiation state diagram enables the effective utilization of multi-round negotiation feedback, significantly improving the timeliness and universality of the generated solution. These will be explained in detail below.

[0033] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for generating non-performing asset disposal plans based on negotiation status, as disclosed in an embodiment of the present invention. Figure 1 The described method for generating non-performing asset disposal plans based on negotiation status can be applied to data processing systems / data processing equipment / data processing servers (wherein, the server includes local processing servers or cloud processing servers). For example... Figure 1 As shown, the method for generating non-performing asset disposal plans based on negotiation status may include the following operations: 101. Obtain the original events between the user and the handling entity in multiple rounds of negotiation, encode the original events into event units, and construct a negotiation event sequence according to the time sequence.

[0034] Optionally, the user can be a debtor who has or may have overdue payments after handling personal loan business within the institution or enterprise acting as the disposing entity.

[0035] Optionally, the disposing entity can be a financial institution or enterprise, which acts as the creditor corresponding to the debtor when the user applies for a personal loan, and promotes the progress of subsequent work of the personal loan business through multiple rounds of negotiation and interaction with the user.

[0036] Optionally, the original event may include one or more of the following: outbound call event, SMS message event, online chat event, self-negotiated page operation event, user callback event, commitment registration event, commitment default event, historical performance event, and human agent remarks event.

[0037] Optionally, the data source of the original event can be a structured field obtained by the handling entity from the business log processing in the business system, or an unstructured field of historical call and / or real-time call transcription text and agent notes, for fields that do not exist in the original event.

[0038] Specifically, the original event is sequentially processed by the encoding module to perform semantic recognition, entity extraction and numerical extraction, thereby obtaining event information such as event timestamp, contact channel type, reporting value, counter-reporting value, response delay, reason for refusal, promised amount, promised date, sentiment tendency, performance result and timeliness requirements, and then the event information is mapped into structured event units.

[0039] Optionally, the event information in the original event corresponds one-to-one with the fields in the structured event unit. Therefore, for the original event, the fields can be filled in by using missing value imputation, confidence labeling, or cross-event aggregation, so as to ensure that the subsequent state recognition and parameter evaluation processes are based on a unified field system.

[0040] As an example, a single event unit can be represented as: ; in, Indicates the event timestamp. Indicates the type of channel to reach. Indicates the current value of the report. This indicates the value returned by the debtor. Indicates response delay. Indicate the reason for rejection. Indicates emotion tags, Indicates the committed amount. Indicates the date of commitment. This indicates the result of contract fulfillment.

[0041] 102. Construct a negotiation state diagram based on the negotiation event sequence, and determine the user's current negotiation state based on the evolution of the negotiation state diagram.

[0042] Optionally, the nodes in the negotiation state graph represent the corresponding negotiation stages in the negotiation event, and the edges of the negotiation state graph represent the state transition relationships between the negotiation stages of each node. The driving factors of the state transition relationships can be one or more of the following: changes in price difference, changes in response delay, changes in sentiment, or changes in the credibility of commitment. This enables the subsequent scheme generation to have a data foundation with event sequence processing and state recognition results.

[0043] Optionally, the negotiation status can be a trial status, a price reduction status, a condition exchange status, a pending transaction status, a delay status, or a low credibility status.

[0044] 103. Based on the boundary constraints of the negotiation state, generate candidate parameter combinations for the disposal plan in the preset parameter search space.

[0045] Optionally, the candidate parameters in the disposal plan may include the plan type z, the reduction / exemption ratio r, the down payment ratio a, the number of installments n, the amount per installment m, and the plan validity period. The parameter combination of the treatment plan can be expressed as: .

[0046] As an example, the solution type z in the disposal plan can be Solution A - a one-time settlement discount solution, in which the number of installments n is fixed at 1; Solution B - a high down payment short installment solution, with a higher down payment ratio a and a shorter installment period n; Solution C - a medium down payment flat installment solution, with the down payment ratio a and the number of installments n falling within the median range of the candidate parameters; and Solution D - a low down payment long installment solution, with a lower down payment ratio a and a longer installment period n. It can be understood that for each parameter in the high, medium, and low classification of the solutions, the mathematical statistics of the corresponding parameters, such as mode, median, standard deviation, and variance, can be calculated through historical multi-round negotiated disposal solutions. Then, the classification boundaries can be set according to business needs and user needs to determine different solution types, which can be flexibly adjusted in actual disposal.

[0047] Optionally, the boundary constraints are state constraint masks generated by the parameter generation module based on the negotiation state, used to limit the upper and lower limits of some parameters under different negotiation state types to constrain the output of candidate parameter combinations.

[0048] Optionally, the parameter search space may include at least the range of reduction / exemption ratios, the range of down payment ratios, the set of installment periods, the range of amounts per period, and the set of validity periods. The parameters in the space may be determined by statistical analysis of historical cases, or the basic parameter space may be determined by regulatory policies, business strategies, and asset classes.

[0049] 104. Based on the preset prediction network, output the acceptance probability and performance probability of each candidate parameter combination, and determine the target parameter combination of the disposal plan according to the acceptance probability and performance probability.

[0050] As can be seen, the above-described embodiments of the invention, by encoding the original events of multiple rounds of negotiation into event units and constructing a negotiation event sequence in chronological order, enable the handling plan to make decisions based on the dynamic evolution trajectory of the user throughout the entire process. This effectively captures the real changing trends of the user's willingness and ability, avoiding misjudgment or missed opportunities due to ignoring intermediate feedback evolution. By abstracting the negotiation event sequence into a negotiation state graph, using nodes and edges in the graph structure to represent different states and their transition relationships in the negotiation process, the implicit characteristics of users scattered in each round of negotiation are uniformly mapped into calculable and evolvable negotiation states. This overcomes the limitations of different agents judging user intentions based on subjective experience, the differences in strategies for similar cases, and the inability to uniformly model them, thus paving the way for subsequent... The solution generation provides standardized state inputs and also serves as a knowledge carrier for negotiation experience across agents and cases. By defining parameter boundaries based on the current negotiation state when generating a solution, multiple candidate solutions are generated within this constraint space. Then, a prediction network is used to estimate the probability of each candidate solution being accepted by the debtor and the probability of the debtor fulfilling its obligations on schedule. This achieves joint optimization of the dual objectives of transaction probability and performance stability, avoiding excessive concessions to facilitate a current transaction that could lead to a high default rate later on, and significantly improving the rationality and controllability of the solution. The evolutionary logic of the negotiation state diagram enables the effective use of feedback from multiple rounds of negotiation, significantly improving the timeliness and universality of the generated solution.

[0051] As an optional embodiment, the step of determining the user's current negotiation state based on the evolution of the negotiation state diagram in the above steps includes: A sliding window is constructed by negotiating event units in a sequence of events. The event units in the negotiation state diagram are statistically analyzed using a sliding window to calculate the mean and slope of change of the driving factors in the state transition relationship, and state identification features are constructed based on the mean and slope of change. The candidate state score is obtained by weighting and summing the state recognition features with state weights and bias terms according to the preset state scoring function. Calculate the candidate state score difference between adjacent sliding windows, and construct a state transition matrix by combining the candidate state score difference and the number of transitions. The scores of each candidate state are smoothed by using the state transition matrix, and the candidate state with the highest score among the candidate states that meet the smoothing criteria is determined as the user's current negotiation state.

[0052] Optionally, when constructing the sliding window, L structured event units can be selected, and the sliding window can be represented as follows: Each event unit in the sliding window includes the reward value, response delay value, emotion tag score, historical commitment fulfillment rate, commitment amount coverage rate, and commitment fulfillment timeliness.

[0053] Optionally, the state recognition features can be represented as: ; in, This represents the state recognition features obtained from the statistics of the k-th sliding window. The mean of the price difference can be represented by the current price value in the event unit. Perform calculations , This represents the average response latency calculated by the k-th sliding window, which can be obtained from the response latency in the event unit. Perform calculations , This represents the average score of the sentiment labels counted in the k-th sliding window, which can be obtained through the sentiment labels of the event unit. The calculations yielded the following results: This represents the average credibility of commitments counted in the k-th sliding window. It can be obtained by comprehensively calculating the historical commitment fulfillment rate, commitment amount coverage rate, and commitment fulfillment timeliness using the commitment amount, commitment date, and fulfillment result marker of the event unit. The slope of change corresponding to the price difference, response delay, sentiment score and commitment credibility in the state recognition features can be used to characterize the changing factors driving state transitions, such as whether the price difference converges, whether the response speed slows down, whether the sentiment tends to stabilize, and whether the commitment credibility decreases.

[0054] Optionally, the state scoring function can be expressed as: ; In the formula, This represents the candidate state score corresponding to the c-th state under the statistics of the k-th sliding window. This represents the state recognition features obtained from the statistics of the k-th sliding window. This represents the state weight corresponding to the c-th state. This represents the bias term corresponding to the c-th state.

[0055] As an example, in the smoothing determination of the state transition matrix, when the price-pressure related characteristics continue to increase and the response delay shortens (determined by the difference and the slope of change), the state is more likely to transition to the price-pressure state. When the debtor proposes that the settlement can be made if the price is reduced by one point, or that it can be accepted if it can be divided into three installments, the state is more likely to transition to the conditional exchange state or the pending transaction state. When the credibility of the commitment decreases and the response delay continues to lengthen, the state is more likely to transition to the delay state or the low credibility state.

[0056] Specifically, the current negotiation state of a user can be determined from the following states: probing, price reduction, condition exchange, pending transaction, delay, or low confidence. This determination can then be used to identify the parameter combination for the subsequent handling plan based on the characteristics of the negotiation state.

[0057] As can be seen, through the above optional embodiments, by dividing the event units in the negotiation state diagram into sliding windows and statistically analyzing the mean and slope of the driving factors within each window, the mean reflects the central tendency of user behavior in the current stage, while the slope characterizes the rate and direction of change in the behavior pattern, thus constituting state recognition features. This enables precise capture of the dynamic evolution trend of debtor behavior characteristics in multiple rounds of negotiation. The state scoring function maps the multi-dimensional state recognition features into quantified candidate state scores, and the state transition matrix smoothly determines candidate states based on the score difference between adjacent windows and historical transition frequency. This ensures that state inference considers both the behavioral characteristics of the current window and is constrained by the historical patterns of state transitions, effectively filtering out temporary state misjudgments caused by emotional fluctuations or data noise in a single dialogue. This system can transform implicit behavioral characteristics from agents' personal experience into a calculable and reproducible state inference process. This allows different agents and different cases to be evaluated in a standardized manner within a unified negotiation state framework, providing a structured technical foundation for the reuse of negotiation experience across cases. By using the highest-scoring candidate state after smoothing the judgment as the sole judgment result for the current negotiation state, it provides a clear stage positioning point for the generation of the handling plan. When user behavior changes substantially, the feature values ​​and scores within the sliding window will evolve accordingly. The state transition matrix drives the smooth migration of the state judgment result to the corresponding stage, thereby naturally triggering the advancement or switching of the plan. This achieves standardized definition of the advancement, maintenance, and switching of negotiation stages in the time dimension, thereby eliminating agent decision differences and improving the efficiency, consistency, and user experience of plan generation.

[0058] As an optional embodiment, the step described above, generating candidate parameter combinations for the disposal scheme in a preset parameter search space based on the boundary constraints of the negotiation state, includes: The basic parameters for determining the parameter search space are obtained from regulatory policies, business strategies, and asset categories. The parameter search space is then constructed based on the actual parameters from historically handled cases. Based on the negotiation status, determine the corresponding scheme type of the disposal plan, and based on the scheme type, determine the parameter boundary of the candidate parameters as the parameter search space, and output the state boundary constraints of the feasible solution corresponding to the candidate parameter combination. Based on the state boundary constraints, feasible solutions for the corresponding parameter combinations of the treatment scheme are obtained by filtering in the parameter search space, forming multiple candidate parameter combinations.

[0059] As an example, when the negotiation is in a trial phase, Option 1 and Option 2 are prioritized for retention, and the reduction / exemption ratio is limited to not exceeding a first threshold. Under conditions of price reduction, Options Two and Three are allowed to enter the candidate set to increase the probability of acceptance within a controllable profit-sharing range; under conditions of exchange or pending transaction, the upper limit of the validity period or the upper limit of the reduction ratio for Options Two and Three can be temporarily expanded, provided that the recovery threshold is met; under conditions of low credibility, Option Four is suppressed, and the down payment ratio is required to be no less than the threshold. And the number of installments does not exceed the threshold. .

[0060] As can be seen, through the above optional embodiments, by determining the basic boundaries of the parameter search space based on regulatory policies, business strategies, and asset classes, it is ensured that the generated candidate solutions always operate within the framework of compliance and the overall risk appetite of the institution. On this basis, the introduction of actual parameters from historical cases ensures that the search space not only has rule constraints but also includes a distribution of effective parameters verified by actual business. This allows candidate solutions to be generated within the same experience space whenever any agent makes a call, fundamentally eliminating inconsistencies caused by differences in individual experience. By directly linking the negotiation status with the solution type, different negotiation statuses correspond to different solution strategy orientations, and corresponding parameter boundary constraints are set accordingly. This enables the disposal solution to automatically match the optimal solution type based on the debtor's current negotiation status and generate candidate combinations within a reasonable parameter range. This ensures that each solution generation is within a feasible domain that matches the user's status, avoiding prematurely proposing aggressive solutions that could lead to a breakdown in negotiations when conditions are not met, and preventing missed opportunities to close deals when the user makes significant concessions. This significantly improves the efficiency and feasibility of solution generation.

[0061] As an optional embodiment, the step above, which outputs the acceptance probability and fulfillment probability for each candidate parameter combination based on a preset prediction network, includes: The specific steps for outputting the acceptance probability are as follows: Extract the difference feature between the report value of the candidate parameter combination and the counter-report value proposed by the user in the previous period; Extract matching features between candidate parameter combinations and users' historical affordable amounts; Extract the acceptance bias feature corresponding to the current negotiation state; The gap feature, matching feature, and acceptance bias feature are used as inputs to the acceptance probability prediction network, and the output is the user's acceptance probability for each candidate parameter combination. The acceptance probability prediction network is trained using a training dataset that includes multiple candidate parameter combinations and original events, as well as a historical case dataset with acceptance probability labels. The specific steps for outputting the fulfillment probability are as follows: The candidate parameter combination and the user's historical performance record are used as input to the performance probability prediction network. The output is the performance probability of the user for each candidate parameter combination. The performance probability prediction network is trained by a training dataset including multiple candidate parameter combinations and original events, as well as a historical case dataset with performance probability labels. The acceptance probability prediction network and the performance probability prediction network were trained together using a public training dataset with different prediction result labels.

[0062] Specifically, the value of the reward for the combination of candidate parameters can be determined based on the reduction ratio r, down payment ratio a, number of installments n, and amount per installment m of the candidate parameters. The value of the reward proposed by the user in the previous period can be determined by the value of the reward for the event information in the original event of multiple rounds of negotiation.

[0063] Specifically, a user's historical affordability amount can be determined based on the maximum value of the reported value in the original event, where the extracted matching features are used to characterize the user's payment ability.

[0064] Optionally, the acceptance probability prediction network can be a three-layer fully connected neural network, with 256 neurons in the first layer, 128 neurons in the second layer, and the third layer outputting the acceptance probability value. This invention does not impose any limitations.

[0065] As an example, the acceptance probability output by the acceptance probability prediction network can be expressed as: ; In the formula, This represents the probability that a user will accept solution x under the given combination of candidate parameters. This represents the Sigmoid function for predicting the acceptance probability of a network. This indicates the difference between the amount of the proposed solution and the value recently reported by the user. This indicates the matching characteristics between the amount of the disposal plan and the user's ability to pay. The bias term representing the current negotiation state. Indicating characteristics of emotional stability, This indicates the timeliness matching characteristics of the solution. As the initial weights, , , , , These are the weight coefficients corresponding to each feature.

[0066] It is understandable that the emotional stability feature and the solution timeliness matching feature in the expression of acceptance probability are alternative conditions for outputting the acceptance probability as an optional implementation method, wherein the emotional stability feature and the solution timeliness matching feature can be determined based on the event units extracted from the original event.

[0067] Specifically, a user's historical performance record can be determined through the performance results of the event information in the original event.

[0068] Optionally, the performance probability prediction network can also be a three-layer fully connected neural network, with 256 neurons in the first layer, 128 neurons in the second layer, and the third layer outputting the performance probability value. This invention does not limit the network.

[0069] As an example, the performance probability prediction network outputs the performance probability as follows: ; In the formula, This represents the probability that the user will fulfill the agreement for the proposed solution x under the given combination of candidate parameters. The sigmoid function represents the network for predicting the probability of fulfillment. Indicates historical performance characteristics, This represents the installment burden characteristics of a user with n installments and m installment amounts. This indicates the average commitment confidence level within the current window. This indicates the characteristics of the risk of delay. These are the initial weighting coefficients. , , , , These are the weighting coefficients corresponding to each feature in the performance probability.

[0070] It is understandable that the installment burden characteristics can be calculated using the number of installments and the installment amount in the candidate parameter combination, the average commitment credibility can be extracted using the status identification characteristics, and the delay risk characteristics can be determined using the negotiation status. The average commitment credibility and delay risk characteristics can be alternative conditions for outputting the performance probability as an optional implementation method.

[0071] As can be seen, through the above optional embodiments, the dynamic difference between the offer value of the candidate parameter combination and the debtor's counteroffer value is extracted by the gap feature, the matching feature captures the fit between the candidate solution and the debtor's historical affordable amount, and the acceptance bias feature reflects the debtor's overall willingness to accept under the current negotiation state. The behavioral evolution of users in multiple rounds of negotiation is fully depicted from the dimensions of bid game, repayment ability, and stage psychology. This enables the prediction network to continuously perceive the dynamic changes in the debtor's attitude, thereby accurately grasping the user's current acceptable range when generating a solution. By sharing a public training dataset containing multiple outcome labels for joint training, the acceptance probability prediction network and the performance probability prediction network can learn from each other and optimize collaboratively in the extraction of underlying features, avoiding the prediction bias and inconsistency that may be caused by independent training. In the inference stage, each candidate parameter combination outputs both the acceptance probability and the performance probability, which transforms the final solution from pursuing a single repayment indicator to optimizing the whole cycle disposal effect. At the same time, the joint training mechanism also enhances the network's generalization ability in data-sparse scenarios and improves the robustness and reliability of the prediction.

[0072] As an optional embodiment, the step of determining the target parameter combination of the disposal plan based on the acceptance probability and the fulfillment probability in the above steps includes: The candidate parameter combinations of the disposal plan are sorted from largest to smallest according to the probability of acceptance and the probability of fulfillment; The acceptance probabilities of the candidate parameter combinations with the highest probability of performance are compared in turn. The candidate parameter combination with the highest probability of acceptance under the condition of the highest probability of performance is selected to obtain the target parameter combination of the disposal plan. Candidate parameter combinations whose probability of performance is no greater than that of the target parameter combination and whose probability of acceptance is no greater than that of the target parameter combination are considered as alternative solutions to the target parameter combination.

[0073] Optionally, the target parameter combination of the disposal plan may include plan identifier, plan type, applicable status, reduction ratio, reduction amount, down payment amount, number of installments, amount per installment, first payment due date, overall validity period, expected acceptance probability, expected performance probability, risk level, and alternative plan identifier.

[0074] As an example, when the disposal plan type is Plan 1, it can correspond to a one-time settlement discount plan and at least include the one-time settlement amount, the corresponding reduction ratio, the effective date and the overdue invalidation rule; when the disposal plan type is Plan 2 or Plan 3, it can further include the down payment amount, the remaining installment plan, the amount of each period, the due date of each period and the return disposal strategy after default; when the system determines that the current negotiation status is a low credibility status, the disposal plan can also forcibly add short time limit and higher down payment requirements.

[0075] In summary, the technical solutions disclosed in the embodiments of the present invention have the following advantages: 1. Improve the alignment between the debt relief plan and the debtor's actual acceptable range; 2. Reduce the time-consuming costs of ineffective price concessions and repeated negotiations; 3. Improve the acceptance rate of negotiations and the rate of contract fulfillment; 4. Improve the consistency of solution outputs across different agents; 5. Enhance the interpretability and auditability of the solution generation process.

[0076] Example 2 Please see Figure 2 , Figure 2 This is a schematic diagram of a non-performing asset disposal plan generation system based on negotiation status, as disclosed in an embodiment of the present invention. Figure 2 The described system for generating non-performing asset disposal plans based on negotiation status can be applied to data processing systems / data processing equipment / data processing servers (wherein, the server includes local processing servers or cloud processing servers). For example... Figure 2 As shown, the system for generating non-performing asset disposal plans based on negotiation status may include: The acquisition module 201 is used to acquire the original events between the user and the handling entity in multiple rounds of negotiation, and encode the original events into event units to construct a negotiation event sequence according to the time sequence. State module 202 is used to construct a negotiation state diagram based on the negotiation event sequence and determine the user's current negotiation state based on the evolution of the negotiation state diagram; The generation module 203 is used to generate candidate parameter combinations for disposal schemes in a preset parameter search space based on the boundary constraints of the negotiation state. The determination module 204 is used to output the acceptance probability and fulfillment probability of each candidate parameter combination based on the preset prediction network, and to determine the target parameter combination of the disposal plan based on the acceptance probability and fulfillment probability.

[0077] As can be seen, the above-described embodiments of the invention, by encoding the original events of multiple rounds of negotiation into event units and constructing a negotiation event sequence in chronological order, enable the handling plan to make decisions based on the dynamic evolution trajectory of the user throughout the entire process. This effectively captures the real changing trends of the user's willingness and ability, avoiding misjudgment or missed opportunities due to ignoring intermediate feedback evolution. By abstracting the negotiation event sequence into a negotiation state graph, using nodes and edges in the graph structure to represent different states and their transition relationships in the negotiation process, the implicit characteristics of users scattered in each round of negotiation are uniformly mapped into calculable and evolvable negotiation states. This overcomes the limitations of different agents judging user intentions based on subjective experience, the differences in strategies for similar cases, and the inability to uniformly model them, thus paving the way for subsequent... The solution generation provides standardized state inputs and also serves as a knowledge carrier for negotiation experience across agents and cases. By defining parameter boundaries based on the current negotiation state when generating a solution, multiple candidate solutions are generated within this constraint space. Then, a prediction network is used to estimate the probability of each candidate solution being accepted by the debtor and the probability of the debtor fulfilling its obligations on schedule. This achieves joint optimization of the dual objectives of transaction probability and performance stability, avoiding excessive concessions to facilitate a current transaction that could lead to a high default rate later on, and significantly improving the rationality and controllability of the solution. The evolutionary logic of the negotiation state diagram enables the effective use of feedback from multiple rounds of negotiation, significantly improving the timeliness and universality of the generated solution.

[0078] As an optional embodiment, the original events include one or more of the following: outbound call events, SMS arrival events, online chat events, self-negotiated page operation events, user callback events, commitment registration events, commitment default events, historical fulfillment events, and human agent remarks events; Negotiation status includes probing status, price reduction status, condition exchange status, pending transaction status, delay status, and low credibility status; The parameter combinations include plan type, reduction / exemption ratio, down payment ratio, number of installments, amount per installment, and plan validity period.

[0079] As an optional embodiment, the original event encoding obtains event information from the original event through semantic recognition, entity extraction, and numerical extraction, and maps the event information into structured event units; The event information includes the event timestamp, contact channel type, reporting value, counter-reporting value, response delay, reason for refusal, promised amount, promised date, sentiment, and performance result.

[0080] As an optional embodiment, the negotiation state diagram includes phase nodes and state transition edges; The stage node is the negotiation stage corresponding to the event unit in the negotiation event sequence; The state transition edge represents the state transition relationship between nodes in each stage during the negotiation stage. The state transition relationship is the driving relationship between one or more of the driving factors such as changes in price difference, changes in response delay, changes in sentiment, or changes in the credibility of commitment and the state changes during the negotiation stage. The changes in the price difference, response delay, sentiment, or credibility of the commitment are determined by the corresponding slope changes calculated in the event information. As an optional embodiment, determining the user's current negotiation state based on the evolution of the negotiation state graph includes: A sliding window is constructed by negotiating event units in a sequence of events. The event units in the negotiation state diagram are statistically analyzed using a sliding window to calculate the mean and slope of change of the driving factors in the state transition relationship, and state identification features are constructed based on the mean and slope of change. The candidate state score is obtained by weighting and summing the state recognition features with state weights and bias terms according to the preset state scoring function. Calculate the candidate state score difference between adjacent sliding windows, and construct a state transition matrix by combining the candidate state score difference and the number of transitions. The scores of each candidate state are smoothed by using the state transition matrix, and the candidate state with the highest score among the candidate states that meet the smoothing criteria is determined as the user's current negotiation state.

[0081] As can be seen, through the above optional embodiments, by dividing the event units in the negotiation state diagram into sliding windows and statistically analyzing the mean and slope of the driving factors within each window, the mean reflects the central tendency of user behavior in the current stage, while the slope characterizes the rate and direction of change in the behavior pattern, thus constituting state recognition features. This enables precise capture of the dynamic evolution trend of debtor behavior characteristics in multiple rounds of negotiation. The state scoring function maps the multi-dimensional state recognition features into quantified candidate state scores, and the state transition matrix smoothly determines candidate states based on the score difference between adjacent windows and historical transition frequency. This ensures that state inference considers both the behavioral characteristics of the current window and is constrained by the historical patterns of state transitions, effectively filtering out temporary state misjudgments caused by emotional fluctuations or data noise in a single dialogue. This system can transform implicit behavioral characteristics from agents' personal experience into a calculable and reproducible state inference process. This allows different agents and different cases to be evaluated in a standardized manner within a unified negotiation state framework, providing a structured technical foundation for the reuse of negotiation experience across cases. By using the highest-scoring candidate state after smoothing the judgment as the sole judgment result for the current negotiation state, it provides a clear stage positioning point for the generation of the handling plan. When user behavior changes substantially, the feature values ​​and scores within the sliding window will evolve accordingly. The state transition matrix drives the smooth migration of the state judgment result to the corresponding stage, thereby naturally triggering the advancement or switching of the plan. This achieves standardized definition of the advancement, maintenance, and switching of negotiation stages in the time dimension, thereby eliminating agent decision differences and improving the efficiency, consistency, and user experience of plan generation.

[0082] As an optional embodiment, based on the boundary constraints of the negotiation state, candidate parameter combinations for the disposal scheme are generated in a preset parameter search space, including: The basic parameters for determining the parameter search space are obtained from regulatory policies, business strategies, and asset categories. The parameter search space is then constructed based on the actual parameters from historically handled cases. Based on the negotiation status, determine the corresponding scheme type of the disposal plan, and based on the scheme type, determine the parameter boundary of the candidate parameters as the parameter search space, and output the state boundary constraints of the feasible solution corresponding to the candidate parameter combination. Based on the state boundary constraints, feasible solutions for the corresponding parameter combinations of the treatment scheme are obtained by filtering in the parameter search space, forming multiple candidate parameter combinations.

[0083] As can be seen, through the above optional embodiments, by determining the basic boundaries of the parameter search space based on regulatory policies, business strategies, and asset classes, it is ensured that the generated candidate solutions always operate within the framework of compliance and the overall risk appetite of the institution. On this basis, the introduction of actual parameters from historical cases ensures that the search space not only has rule constraints but also includes a distribution of effective parameters verified by actual business. This allows candidate solutions to be generated within the same experience space whenever any agent makes a call, fundamentally eliminating inconsistencies caused by differences in individual experience. By directly linking the negotiation status with the solution type, different negotiation statuses correspond to different solution strategy orientations, and corresponding parameter boundary constraints are set accordingly. This enables the disposal solution to automatically match the optimal solution type based on the debtor's current negotiation status and generate candidate combinations within a reasonable parameter range. This ensures that each solution generation is within a feasible domain that matches the user's status, avoiding prematurely proposing aggressive solutions that could lead to a breakdown in negotiations when conditions are not met, and preventing missed opportunities to close deals when the user makes significant concessions. This significantly improves the efficiency and feasibility of solution generation.

[0084] As an optional embodiment, the acceptance probability and fulfillment probability of each candidate parameter combination are output based on a preset prediction network, including: The specific steps for outputting the acceptance probability are as follows: Extract the difference feature between the report value of the candidate parameter combination and the counter-report value proposed by the user in the previous period; Extract matching features between candidate parameter combinations and users' historical affordable amounts; Extract the acceptance bias feature corresponding to the current negotiation state; The gap feature, matching feature, and acceptance bias feature are used as inputs to the acceptance probability prediction network, and the output is the user's acceptance probability for each candidate parameter combination. The acceptance probability prediction network is trained using a training dataset that includes multiple candidate parameter combinations and original events, as well as a historical case dataset with acceptance probability labels. The specific steps for outputting the fulfillment probability are as follows: The candidate parameter combination and the user's historical performance record are used as input to the performance probability prediction network. The output is the performance probability of the user for each candidate parameter combination. The performance probability prediction network is trained by a training dataset including multiple candidate parameter combinations and original events, as well as a historical case dataset with performance probability labels. The acceptance probability prediction network and the performance probability prediction network were trained together using a public training dataset with different prediction result labels.

[0085] As can be seen, through the above optional embodiments, the dynamic difference between the offer value of the candidate parameter combination and the debtor's counteroffer value is extracted by the gap feature, the matching feature captures the fit between the candidate solution and the debtor's historical affordable amount, and the acceptance bias feature reflects the debtor's overall willingness to accept under the current negotiation state. The behavioral evolution of users in multiple rounds of negotiation is fully depicted from the dimensions of bid game, repayment ability, and stage psychology. This enables the prediction network to continuously perceive the dynamic changes in the debtor's attitude, thereby accurately grasping the user's current acceptable range when generating a solution. By sharing a public training dataset containing multiple outcome labels for joint training, the acceptance probability prediction network and the performance probability prediction network can learn from each other and optimize collaboratively in the extraction of underlying features, avoiding the prediction bias and inconsistency that may be caused by independent training. In the inference stage, each candidate parameter combination outputs both the acceptance probability and the performance probability, which transforms the final solution from pursuing a single repayment indicator to optimizing the whole cycle disposal effect. At the same time, the joint training mechanism also enhances the network's generalization ability in data-sparse scenarios and improves the robustness and reliability of the prediction.

[0086] As an optional embodiment, the target parameter combination for the disposal plan is determined based on the acceptance probability and the fulfillment probability, including: The candidate parameter combinations of the disposal plan are sorted from largest to smallest according to the probability of acceptance and the probability of fulfillment; The acceptance probabilities of the candidate parameter combinations with the highest probability of performance are compared in turn. The candidate parameter combination with the highest probability of acceptance under the condition of the highest probability of performance is selected to obtain the target parameter combination of the disposal plan. Candidate parameter combinations whose probability of performance is no greater than that of the target parameter combination and whose probability of acceptance is no greater than that of the target parameter combination are considered as alternative solutions to the target parameter combination.

[0087] Example 3 Please see Figure 3 , Figure 3 This is another non-performing asset disposal scheme generation system based on negotiation status disclosed in the embodiments of the present invention. Figure 3 The described system for generating non-performing asset disposal plans based on negotiation status is applied in a data processing system / data processing equipment / data processing server (wherein, the server includes a local processing server or a cloud processing server). For example... Figure 3 As shown, the system for generating non-performing asset disposal plans based on negotiation status may include: Memory 301 storing executable program code; Processor 302 coupled to memory 301; The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the method for generating a non-performing asset disposal plan based on the negotiation state described in Embodiment 1.

[0088] Example 4 This invention discloses a computer read storage medium that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the steps of the method for generating a non-performing asset disposal plan based on a negotiated state as described in Embodiment 1.

[0089] Example 5 This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of the method for generating a non-performing asset disposal plan based on a negotiation state as described in Embodiment 1.

[0090] The foregoing has described specific embodiments of this specification; other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily have to follow the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0091] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0092] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0093] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0094] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0095] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0096] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0097] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0098] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0099] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0100] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0101] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0102] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0103] Finally, it should be noted that the method and system for generating non-performing asset disposal plans based on negotiation status disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating non-performing asset disposal plans based on negotiation status, characterized in that, The method includes: The original events between the user and the handling entity during multiple rounds of negotiation are obtained, and the original events are encoded into event units to construct a negotiation event sequence in chronological order. A negotiation state diagram is constructed based on the negotiation event sequence, and the user's current negotiation state is determined based on the evolution of the negotiation state diagram. Based on the boundary constraints of the negotiation state, candidate parameter combinations for the disposal plan are generated in the preset parameter search space. Based on the preset prediction network, the acceptance probability and fulfillment probability of each candidate parameter combination are output, and the target parameter combination of the disposal plan is determined according to the acceptance probability and fulfillment probability.

2. The method for generating non-performing asset disposal plans based on negotiation status according to claim 1, characterized in that, The original events include one or more of the following: outbound call events, SMS delivery events, online chat events, self-negotiated page operation events, user callback events, commitment registration events, commitment default events, historical fulfillment events, and human agent remarks events; Negotiation status includes probing status, price reduction status, condition exchange status, pending transaction status, delay status, and low credibility status; The parameter combinations include plan type, reduction / exemption ratio, down payment ratio, number of installments, amount per installment, and plan validity period.

3. The method for generating non-performing asset disposal plans based on negotiation status according to claim 2, characterized in that, The original event encoding obtains event information from the original event through semantic recognition, entity extraction, and numerical extraction, and maps the event information into structured event units; The event information includes the event timestamp, contact channel type, reporting value, counter-reporting value, response delay, reason for refusal, promised amount, promised date, sentiment, and performance result.

4. The method for generating non-performing asset disposal plans based on negotiation status according to claim 3, characterized in that, The negotiation state diagram includes phase nodes and state transition edges; The stage node is the negotiation stage corresponding to the event unit in the negotiation event sequence; The state transition edge represents the state transition relationship between nodes in each stage during the negotiation stage. The state transition relationship is the driving relationship between one or more of the driving factors such as changes in price difference, changes in response delay, changes in sentiment, or changes in the credibility of commitment and the state changes during the negotiation stage. The changes in the price difference, response delay, sentiment, or credibility of the commitment are determined by the corresponding slope changes calculated in the event information.

5. The method for generating non-performing asset disposal plans based on negotiation status according to claim 1, characterized in that, The user's current negotiation state is determined based on the evolution of the negotiation state diagram, including: A sliding window is constructed by negotiating event units in a sequence of events. The event units in the negotiation state diagram are statistically analyzed using a sliding window to calculate the mean and slope of change of the driving factors in the state transition relationship, and state identification features are constructed based on the mean and slope of change. The candidate state score is obtained by weighting and summing the state recognition features with state weights and bias terms according to the preset state scoring function. Calculate the candidate state score difference between adjacent sliding windows, and construct a state transition matrix by combining the candidate state score difference and the number of transitions. The scores of each candidate state are smoothed by using the state transition matrix, and the candidate state with the highest score among the candidate states that meet the smoothing criteria is determined as the user's current negotiation state.

6. The method for generating non-performing asset disposal plans based on negotiation status according to claim 1, characterized in that, Based on the boundary constraints of the negotiated state, candidate parameter combinations for the disposal plan are generated in a pre-defined parameter search space, including: The basic parameters for determining the parameter search space are obtained from regulatory policies, business strategies, and asset categories. The parameter search space is then constructed based on the actual parameters from historically handled cases. Based on the negotiation status, determine the corresponding scheme type of the disposal plan, and based on the scheme type, determine the parameter boundary of the candidate parameters as the parameter search space, and output the state boundary constraints of the feasible solution corresponding to the candidate parameter combination. Based on the state boundary constraints, feasible solutions for the corresponding parameter combinations of the treatment scheme are obtained by filtering in the parameter search space, forming multiple candidate parameter combinations.

7. The method for generating non-performing asset disposal plans based on negotiation status according to claim 1, characterized in that, Based on the pre-defined prediction network, the acceptance probability and fulfillment probability of each candidate parameter combination are output, including: The specific steps for outputting the acceptance probability are as follows: Extract the difference feature between the report value of the candidate parameter combination and the counter-report value proposed by the user in the previous period; Extract matching features between candidate parameter combinations and users' historical affordable amounts; Extract the acceptance bias feature corresponding to the current negotiation state; The gap feature, matching feature, and acceptance bias feature are used as inputs to the acceptance probability prediction network, and the output is the user's acceptance probability for each candidate parameter combination. The acceptance probability prediction network is trained using a training dataset that includes multiple candidate parameter combinations and original events, as well as a historical case dataset with acceptance probability labels. The specific steps for outputting the fulfillment probability are as follows: The candidate parameter combination and the user's historical performance record are used as input to the performance probability prediction network. The output is the performance probability of the user for each candidate parameter combination. The performance probability prediction network is trained by a training dataset including multiple candidate parameter combinations and original events, as well as a historical case dataset with performance probability labels. The acceptance probability prediction network and the performance probability prediction network were trained together using a public training dataset with different prediction result labels.

8. The method for generating non-performing asset disposal plans based on negotiation status according to claim 1, characterized in that, The target parameter combination for the disposal plan is determined based on the acceptance probability and the performance probability, including: The candidate parameter combinations of the disposal plan are sorted from largest to smallest according to the probability of acceptance and the probability of fulfillment; The acceptance probabilities of the candidate parameter combinations with the highest probability of performance are compared in turn. The candidate parameter combination with the highest probability of acceptance under the condition of the highest probability of performance is selected to obtain the target parameter combination of the disposal plan. Candidate parameter combinations whose probability of performance is no greater than that of the target parameter combination and whose probability of acceptance is no greater than that of the target parameter combination are considered as alternative solutions to the target parameter combination.

9. A system for generating non-performing asset disposal plans based on negotiation status, characterized in that, The system includes: The acquisition module is used to acquire the original events between the user and the handling entity in multiple rounds of negotiation, encode the original events into event units, and construct a negotiation event sequence according to the time sequence. The state module is used to construct a negotiation state diagram based on the negotiation event sequence, and to determine the user's current negotiation state based on the evolution of the negotiation state diagram. The generation module is used to generate candidate parameter combinations for disposal schemes in a preset parameter search space based on the boundary constraints of the negotiation state. The determination module is used to output the acceptance probability and fulfillment probability of each candidate parameter combination based on the preset prediction network, and to determine the target parameter combination of the disposal plan based on the acceptance probability and fulfillment probability.

10. A system for generating non-performing asset disposal plans based on negotiation status, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method for generating a non-performing asset disposal plan based on a negotiation state as described in any one of claims 1-8.

Images