Multi-objective optimized home decoration consumption staging product multi-party collaborative configuration method and system

Abstract

The invention relates to a multi-objective optimized home decoration consumption installment product multi-party collaborative configuration method and system, and the method comprises the following steps: S1, obtaining multi-source data related to installment product configuration, and processing the multi-source data to form a modeling data set; s2, based on the modeling data set, respectively constructing three types of target models oriented to consumers, merchants and resource parties; s3, establishing a multi-objective optimization model taking the three types of objective models as objective functions, and solving through candidate solution constraint pre-screening and a non-dominated solution increment maintenance mechanism to obtain a Pareto optimal scheme set; s4, the obtained Pareto optimal scheme set is sorted, screened and output, and staging product configuration parameters are formed; s5, applying the staging product configuration parameters to product delivery, and monitoring a delivery effect and post-service performance; according to the method, the three-party target model is constructed, the solving efficiency is improved through constraint pre-screening and solution set increment maintenance, and the configuration parameters are output in the calculation power or time delay prediction.

Description

Technical Field

[0001] This invention relates to the field of computer implementation technology for data processing and optimization decision-making, and in particular to a multi-objective optimization method and system for multi-party collaborative configuration of installment payment products for home decoration consumption. Background Technology

[0002] The selection of installment payment products for home renovation needs to consider multiple factors simultaneously, including consumer acceptance, merchant conversion and fulfillment performance, and resource provider costs and risk-return considerations. Furthermore, it is subject to compliance, risk thresholds, and operational rules. Existing methods often rely on experience or a single objective, making it difficult to achieve a stable balance among multiple goals and easily leading to fluctuations in the effectiveness of the offerings.

[0003] Meanwhile, the diverse sources of relevant data, inconsistent definitions and time series, and frequent omissions and anomalies lead to poor model output stability and reliability without effective alignment and consistency processing. Furthermore, multi-objective solutions incur high computational costs when facing numerous constraints and large candidate spaces, making it difficult to meet online latency requirements. The lack of triggering, verification, and rollback controls for post-deployment feedback updates can easily result in frequent adjustments and untraceability issues. Therefore, an efficient, stable, and auditable computer implementation solution is needed. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this invention is to provide a multi-objective optimized multi-party collaborative configuration method and system for home decoration consumer installment products, so as to solve one or more problems in the prior art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: A multi-objective optimization method for home improvement installment payment products involving multi-party collaboration includes the following steps: S1. Acquire multi-source data: Acquire multi-source data related to installment product configuration. Multi-source data includes data from consumers, merchants, and resource providers. Process the multi-source data to form a modeling dataset. S2. Construct three types of target models: Based on the modeling dataset, construct three types of target models for consumers, merchants, and resource providers respectively. The three types of target models include: consumer choice response model, merchant conversion revenue model, resource provider default or failure risk estimation model, and risk-adjusted revenue model. S3. Solving for the Pareto optimal solution set: Under the constraints of risk threshold, parameter range, and system computing resources or response delay, a multi-objective optimization model with three types of objective models as objective functions is established, and the Pareto optimal solution set is obtained by pre-screening of candidate solutions and incremental maintenance mechanism of non-dominated solutions. S4. Forming phased product configuration parameters: Sort, filter and output the obtained Pareto optimal solution set to form phased product configuration parameters, and simultaneously output the constraint satisfaction record, model version information and filtering strategy identifier corresponding to the configuration parameters. S5. Update installment product configuration parameters: Apply the installment product configuration parameters to product launch and monitor the launch effect and post-service performance; when the monitoring results meet the preset update trigger conditions, perform controlled feedback updates on the three target models and installment product configuration parameters to achieve dynamic iterative optimization of installment product configuration and output stability control.

[0006] The above solution acquires relevant data from consumers, merchants, and resource providers, performs time alignment, handles missing and anomalies, and verifies consistency. It then constructs target models for consumer response, merchant revenue, and resource provider risk / return. Under risk thresholds and business rule constraints, it performs multi-objective optimization, using constraint pre-screening and incremental maintenance of non-dominated solutions to generate and filter Pareto solution sets. It outputs parameters such as the number of installments, down payment ratio, fee structure, and interest subsidy allocation, and generates constraint satisfaction records and model version information. When trigger conditions are met, it performs controlled updates and supports verification and rollback. This solves the problems of inconsistent multi-source data, high solution overhead, and unstable online updates in home renovation installment configuration. Furthermore, under preset latency or computing power constraints, it enables multi-party collaborative configuration, improving solution efficiency and output stability.

[0007] Furthermore, multi-source data includes consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, resource provider cost data and post-service performance data, among which: Consumer credit information includes one or more of the following: consumer credit score, debt level, historical overdue records, income, and stability indicators; Consumer payment capacity data includes one or more of the following: consumer disposable cash flow, monthly spending structure, repayment affordability, and payment preferences; Merchant transaction characteristic data includes one or more of the following: average order value distribution, conversion rate, customer complaint rate, refund rate, order cancellation rate, and product category structure. Merchant performance-related data includes one or more of the following: installation performance time, delivery completion rate, delayed performance rate, after-sales processing time, dispute resolution results, refund completion time, and performance exception records. Resource provider cost data includes one or more of the following: resource occupancy costs, channel costs, and operating costs. The performance data of the resource provider after providing services includes one or more of the following: delinquency rate, default rate, recovery rate, loss rate, and migration rate.

[0008] Furthermore, the processing of multi-source data includes time window alignment, latency compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification, among which: After performing time window alignment and delay compensation on multi-source data, samples are obtained, and a data quality score is generated for each sample. The data quality score is determined by the missing proportion, outlier proportion, and consistency check results. After cleaning, handling missing values, detecting anomalies, constructing features, and standardizing the multi-source data, a dataset for modeling is formed. Dataset The expression is: , in, This represents the consumer feature vector corresponding to the i-th sample. This represents the merchant feature vector corresponding to the i-th sample. This represents the resource-side feature vector corresponding to the i-th sample. This represents the historical result label or statistical result corresponding to the i-th sample. This represents the data quality score corresponding to the i-th sample. This represents the total number of samples.

[0009] Furthermore, consumer credit information and payment ability data are input into the consumer choice response model construction process to establish a probability function or acceptance function for consumers' choice of different installment product configuration parameters. It outputs consumer target indicators associated with the installment product configuration parameters as the consumer objective function. ; In the process of constructing a consumer choice response model, the obtained data quality score is used as a basis. The samples are weighted to reduce the impact of low-quality data on model stability; Choose either probability or acceptability function satisfy: ,and , in, This represents the Sigmoid function. Represents the feature mapping function. This represents the consumer's choice of response model parameters, where T represents the transpose and x represents the installment product configuration parameter vector.

[0010] Furthermore, a merchant conversion revenue function is established based on merchant transaction characteristic data. And introduce penalties for refunds or disputes. Adjusting the returns, we obtain the merchant's objective function. And based on the obtained data quality score Weight the training samples or statistical estimates; Merchant conversion revenue function The expression is: , Merchant objective function The expression is: , in, This represents the merchant conversion revenue mapping function. This represents the parameters of the merchant conversion revenue model. Indicates the penalty coefficient. .

[0011] Furthermore, based on resource provider cost data and post-service performance data, a risk probability estimation model and a loss assessment model are established to output the risk probability. and construct the expected loss In the expected loss Based on this, a risk-adjusted return function is constructed as the resource-side objective function. ; Output risk probability The expression is: ; Expected loss The expression is: ; Resource side objective function The expression is: ; in, Represents the feature mapping function. This represents the parameters of the resource provider default or failure risk estimation model. This represents the feature vector of the resource provider. The loss rate function represents the percentage of loss in the event of default or failure. This represents the risk exposure function, used to characterize the exposure balance or exposure amount in the event of default or failure, where Π(x) represents the expected return term. This includes resource costs and operating costs.

[0012] Furthermore, risk threshold constraints and business rule constraints include risk thresholds. Risk-adjusted return lower limit threshold The upper and lower bounds of the allowable range for rate structure parameters are respectively and The upper and lower limits of the permissible down payment ratio are respectively and and the optional set of periods ; in, , , , ; Installment Product Configuration Parameter Vector The expression is: , Constructing multi-objective function vectors Multi-objective function vector The expression is: , in, Indicates the period number. , Indicates the down payment percentage. Indicates the rate structure parameters, Indicates the interest subsidy sharing parameter; The interest subsidy sharing parameter s includes the consumer's interest subsidy sharing ratio. Merchant interest subsidy sharing ratio Interest subsidy sharing ratio with resource providers , .

[0013] Furthermore, S3 includes the following steps: S3-1. Under the constraint set Ω consisting of risk threshold constraint and parameter range constraint, perform constraint pre-screening on candidate solutions and eliminate candidate solutions that do not satisfy the constraint set Ω. S3-2. Calculate the multi-objective function vector for the pre-screened candidate solutions. ; S3-3. Generate a set of non-dominated solutions using Pareto dominance relations, and incrementally maintain the set of non-dominated solutions. For any two sets of candidate solutions... If the following conditions are met: , , but And the solutions that are not dominated by any other solution form a Pareto optimal solution set; S3-4. Under preset computing resource or response delay constraints, output the top K non-dominated solutions of the Pareto optimal solution set or the solution set that satisfies the coverage threshold. S3-5. Using a weighted scoring function Sort or make a quadratic decision on the set of Pareto optimal solutions, using a weighted scoring function. The expression is: , and , in, Configure parameters for the target weights, which correspond to the consumer objective function. Merchant objective function and resource objective function ; S3-6. Select the weighted scoring function. The output of the scheme with the largest value is the phased product configuration parameter combination, and an audit record is generated simultaneously for this output scheme. This audit record includes the selected scheme. Corresponding objective function value Constraint satisfaction results, model version number, and filtering strategy identifier.

[0014] Furthermore, after the installment product was launched, indicators such as application approval rate, conversion rate, refund rate, delinquency rate, default rate, and risk-adjusted return were collected to form an effectiveness evaluation vector Yt, which was then processed to obtain the new dataset. Based on the new dataset Model parameter sets for three types of target models Update; The update is executed when a preset update trigger condition is met. The update trigger condition includes any of the following: effect evaluation vector. The model update satisfies the following conditions: the offset from the historical baseline exceeds a threshold, the drift of the input feature distribution exceeds a threshold, or the number of new samples reaches a threshold. ,in For learning rate, For loss function, , This indicates the new dataset used for model updates; If the stability metrics and constraint satisfaction rate do not meet the preset standards in the validation set or sliding time window after the update, the model version before the update will be rolled back. Simultaneously, the range or candidate set of values ​​for the number of periods, down payment ratio, fee structure, and interest subsidy allocation parameters are adjusted to reduce parameter oscillations, improve constraint satisfaction rate, and enhance system output stability.

[0015] A multi-objective collaborative optimization system for home decoration consumer installment products, employing any of the methods described above, includes: a multi-source data acquisition and processing module, used to acquire and process consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, resource provider cost data and post-service performance data, and to perform time window alignment, delay compensation, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification to form a modeling dataset; The three-party target modeling module is used to construct consumer choice response models, merchant conversion revenue models, resource provider risk estimation models, and risk-adjusted revenue models, and outputs the results. , , ; The multi-objective optimization module is used to establish a multi-objective optimization model and generate a Pareto optimal solution set under risk thresholds and business rule constraints, wherein the following conditions are met: Furthermore, by using a constraint pre-screening and non-dominated solution incremental maintenance mechanism, a solution set is output under preset computing resource or response delay constraints, in order to reduce the candidate solution search space, reduce the number of invalid evaluations, and improve the system's solution efficiency. The scheme selection and parameter output configuration module is used to select the Pareto optimal scheme set and output the number of periods, down payment ratio, rate structure and interest subsidy allocation configuration parameters, and generate audit records containing constraint satisfaction records and model version information. The monitoring and feedback update module is used to monitor the effectiveness of the campaign and post-service performance, and to update based on preset trigger conditions. Alternatively, an equivalent update rule can be used to update the model, and if the update fails verification, a version rollback can be performed. At the same time, the candidate set or value range of configuration parameters can be adjusted to reduce output fluctuations caused by model update failures and improve system stability.

[0016] Compared with the prior art, the beneficial technical effects of the present invention are as follows: (i) This invention solves the technical problems in the prior art that lead to modeling input distortion and large result fluctuations due to inconsistent caliber, inconsistent time sequence, and many missing anomalies in multi-source data by uniformly acquiring and processing multi-source heterogeneous data from consumers, merchants, and resource providers, and further performing time window alignment, delay compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification. At the same time, by generating data quality scores and using them for weighted control in model training or statistical estimation, the interference of low-quality samples on model parameter learning can be reduced, thereby improving the consistency, effectiveness, and usability of modeling data, and enhancing the stability, accuracy, and reliability of model training and application stages.

[0017] (II) This invention constructs a consumer choice response model, a merchant conversion revenue model, and a resource provider default or failure risk estimation model and risk-adjusted revenue model, respectively, forming a multi-objective function system for consumers, merchants, and resource providers. This solves the technical problems in the prior art where parameter configuration is based solely on a single revenue objective or empirical rules, making it difficult to balance consumer acceptance, merchant conversion effect, and risk-return for resource providers. By incorporating the objectives of the three parties into a unified calculation framework for collaborative modeling and joint optimization, the phased product configuration parameters can achieve a more reasonable balance among the interests of multiple parties, avoiding local optima that lead to a decline in overall effect, thereby improving the scientific nature, coordination, and practical adaptability of the output solution.

[0018] (III) This invention solves the technical problems of excessively large candidate parameter space, excessive invalid evaluation, high computational overhead, and insufficient online response capability in the multi-objective solution process by setting risk threshold constraints, parameter range constraints, and system computing resource or response delay constraints, and by combining candidate solution constraint pre-screening, non-dominated solution incremental maintenance, and Pareto optimal solution set output mechanism. In particular, by pre-eliminating candidate solutions that do not meet the constraints, the search space can be effectively compressed. By incrementally maintaining the non-dominated solution set, repeated comparisons and invalid calculations can be reduced, thereby reducing system resource consumption, improving the efficiency of multi-objective optimization solution and real-time response capability, and making the scheme more suitable for online deployment and stable operation in actual computer systems.

[0019] (iv) This invention sorts, filters, and outputs the Pareto optimal solution set, and simultaneously generates constraint satisfaction records, model version information, and filtering strategy identifiers. It also combines deployment effect monitoring, update trigger judgment, controlled feedback update, verification failure rollback, and candidate parameter set adjustment mechanisms to solve the technical problems in the prior art where optimization results can be generated but lack clear filtering criteria, are difficult to audit and trace, and the model update process is prone to parameter oscillation, output instability, and difficulty in timely recovery after anomalies. Therefore, this invention can not only achieve interpretable output, version traceability, and process auditability of phased product configuration parameters, but also perform controlled iteration based on feedback during operation and quickly recover to a stable state when update anomalies occur, thereby improving the long-term stability, controllability, and engineering application value of the system. Attached Figure Description

[0020] Figure 1 The flowchart illustrates a multi-objective optimization method and system for multi-party collaborative configuration of home decoration installment products according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the device proposed by this invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of this invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, only for the purpose of conveniently and clearly illustrating the embodiments of this invention. Please refer to the drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the implementation conditions of this invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportional relationships, or adjustments to the size, without affecting the effects and objectives achieved by this invention, should still fall within the scope of the technical content disclosed in this invention.

[0022] Accordingly, this invention also provides a multi-objective optimized multi-party collaborative configuration method for home decoration installment products, including the following steps: S1. Acquire multi-source data: Acquire multi-source data related to installment product configuration. Multi-source data includes data from consumers, merchants, and resource providers. Process the multi-source data to form a modeling dataset. S2. Construct three types of target models: Based on the modeling dataset, construct three types of target models for consumers, merchants, and resource providers respectively. The three types of target models include: consumer choice response model, merchant conversion revenue model, resource provider default or failure risk estimation model, and risk-adjusted revenue model. S3. Solving for the Pareto optimal solution set: Under the constraints of risk threshold, parameter range, and system computing resources or response delay, a multi-objective optimization model with three types of objective models as objective functions is established, and the Pareto optimal solution set is obtained by pre-screening of candidate solutions and incremental maintenance mechanism of non-dominated solutions. S4. Forming phased product configuration parameters: Sort, filter and output the obtained Pareto optimal solution set to form phased product configuration parameters, and simultaneously output the constraint satisfaction record, model version information and filtering strategy identifier corresponding to the configuration parameters. S5. Update installment product configuration parameters: Apply the installment product configuration parameters to product launch and monitor the launch effect and post-service performance; when the monitoring results meet the preset update trigger conditions, perform controlled feedback updates on the three target models and installment product configuration parameters to achieve dynamic iterative optimization of installment product configuration and output stability control.

[0023] The above solution acquires relevant data from consumers, merchants, and resource providers, performs time alignment, handles missing and anomalies, and verifies consistency. It then constructs target models for consumer response, merchant revenue, and resource provider risk / return. Under risk thresholds and business rule constraints, it performs multi-objective optimization, using constraint pre-screening and incremental maintenance of non-dominated solutions to generate and filter Pareto solution sets. It outputs parameters such as the number of installments, down payment ratio, fee structure, and interest subsidy allocation, and generates constraint satisfaction records and model version information. When trigger conditions are met, it performs controlled updates and supports verification and rollback. This solves the problems of inconsistent multi-source data, high solution overhead, and unstable online updates in home renovation installment configuration. Furthermore, under preset latency or computing power constraints, it enables multi-party collaborative configuration, improving solution efficiency and output stability.

[0024] Furthermore, multi-source data includes consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, resource provider cost data and post-service performance data, among which: Consumer credit information includes one or more of the following: consumer credit score, debt level, historical overdue records, income, and stability indicators; Consumer payment capacity data includes one or more of the following: consumer disposable cash flow, monthly spending structure, repayment affordability, and payment preferences; Merchant transaction characteristic data includes one or more of the following: average order value distribution, conversion rate, customer complaint rate, refund rate, order cancellation rate, and product category structure. Merchant performance-related data includes one or more of the following: installation performance time, delivery completion rate, delayed performance rate, after-sales processing time, dispute resolution results, refund completion time, and performance exception records. Resource provider cost data includes one or more of the following: resource occupancy costs, channel costs, and operating costs. The performance data of the resource provider after providing services includes one or more of the following: delinquency rate, default rate, recovery rate, loss rate, and migration rate.

[0025] Furthermore, the processing of multi-source data includes time window alignment, latency compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification, among which: After performing time window alignment and delay compensation on multi-source data, samples are obtained, and a data quality score is generated for each sample. The data quality score is determined by the missing proportion, outlier proportion, and consistency check results.

[0026] Specifically, consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, and resource provider cost data and post-service performance data are aligned with the observation window corresponding to the unified decision-making time point. Delay compensation processing is performed on data with different collection time, storage time, or effective time to form a time-consistent joint sample data composed of consumer characteristics, merchant characteristics, and resource provider characteristics.

[0027] After cleaning, handling missing values, detecting anomalies, constructing features, and standardizing the multi-source data, a dataset for modeling is formed. Dataset The expression is: , in, This represents the consumer feature vector corresponding to the i-th sample. This represents the merchant feature vector corresponding to the i-th sample. This represents the resource-side feature vector corresponding to the i-th sample. This represents the historical result label or statistical result corresponding to the i-th sample. This represents the data quality score corresponding to the i-th sample. This represents the total number of samples.

[0028] Specifically, the multi-source data after time window alignment and delay compensation is cleaned, missing value processing is performed, and anomaly detection is performed. Cleaning includes deduplication, format correction, field encoding standardization, unit of measurement standardization, and removal of illegal records. Missing value processing includes using one or more of the following methods to complete the missing values: mean, median, mode, stratified statistical values, preset missing value markers, or model estimates. Anomaly detection includes identifying outliers based on business thresholds, statistical distribution intervals, historical fluctuation ranges, or logical constraints, and performing correction, truncation, weight reduction, or removal processing on outliers.

[0029] The processed multi-source data undergoes feature construction and standardization. Feature construction includes generating consumer solvency features, merchant conversion and fulfillment features, resource provider risk exposure and cost features, and cross features related to installment product configuration parameters. Standardization includes one or more of the following methods: Z-score standardization, interval normalization, logarithmic transformation, or other scale-unified processing methods.

[0030] The standardized features are subjected to feature consistency verification, which includes field type consistency verification, value range consistency verification, primary key association consistency verification, business logic consistency verification, and feature processing rule consistency verification between the training and application phases. Based on the missing ratio, abnormal field ratio, and consistency verification results, a data quality score is generated for each sample.

[0031] in, This represents the consumer feature vector corresponding to the i-th sample, which is formed by aligning consumer credit information and payment ability data through time windows, delay compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification. This represents the merchant feature vector corresponding to the i-th sample, which is formed by aligning merchant transaction features and performance-related data through time windows, delay compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification. This represents the resource provider feature vector corresponding to the i-th sample, which is formed by aligning resource provider cost data and post-service performance data through time window, delay compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification. This represents the historical result label or statistical result corresponding to the i-th sample; This represents the data quality score generated for the i-th sample based on the missing proportion, the proportion of outliers, and the feature consistency verification results.

[0032] Furthermore, consumer credit information and payment ability data are input into the consumer choice response model construction process to establish a probability function or acceptance function for consumers' choice of different installment product configuration parameters. It outputs consumer target indicators associated with the installment product configuration parameters as the consumer objective function. ; In the process of constructing a consumer choice response model, the obtained data quality score is used as a basis. The samples are weighted to reduce the impact of low-quality data on model stability; Choose either probability or acceptability function satisfy: ,and , in, This represents the Sigmoid function. Represents the feature mapping function. This represents the consumer's choice of response model parameters, where T represents the transpose and x represents the installment product configuration parameter vector.

[0033] Specifically, the consumer feature vector corresponding to the i-th sample... By inputting the installment product configuration parameter vector x into the consumer choice response model construction process, a probability or acceptance function for consumers' choice of different installment product configuration parameters is established. It outputs consumer target indicators associated with installment product configuration parameters based on the selection probability or acceptability function, serving as the consumer target function. .

[0034] In constructing a consumer choice response model, the data quality score corresponding to the i-th sample is used as a basis. Weighting the sample loss term or statistical estimation term can reduce the impact of samples with high missing rates, many outliers, or poor feature consistency verification results on the learning of parameters in the consumer choice response model, thereby reducing the impact of low-quality data on the stability of the consumer choice response model.

[0035] Furthermore, a merchant conversion revenue function is established based on merchant transaction characteristic data. And introduce penalties for refunds or disputes. Adjusting the returns, we obtain the merchant's objective function. And based on the obtained data quality score Weight the training samples or statistical estimates; Merchant conversion revenue function The expression is: , Merchant objective function The expression is: , in, This represents the merchant conversion revenue mapping function. This represents the parameters of the merchant conversion revenue model. Indicates the penalty coefficient. .

[0036] Specifically, the merchant feature vector corresponding to the i-th sample... By inputting the installment product configuration parameter vector x into the merchant conversion revenue model construction process, a merchant conversion revenue function is established to characterize the comprehensive result of merchant-side transaction conversion and fulfillment costs under candidate configuration parameters. The merchant conversion revenue function serves as the computational input in the multi-objective optimization module, participating in the candidate solution selection. During the construction of the merchant conversion revenue model, the data quality score corresponding to the i-th sample is used. Weighting is applied to sample loss terms or statistical estimates to reduce the impact of samples with high missing rates, many outliers, or poor feature consistency verification results on the learning of merchant conversion revenue model parameters or revenue statistical estimation.

[0037] Furthermore, based on resource provider cost data and post-service performance data, a risk probability estimation model and a loss assessment model are established to output the risk probability. and construct the expected loss In the expected loss Based on this, a risk-adjusted return function is constructed as the resource-side objective function. ; Output risk probability The expression is: ; Expected loss The expression is: ; Resource side objective function The expression is: ; in, Represents the feature mapping function. This represents the parameters of the resource provider default or failure risk estimation model. This represents the feature vector of the resource provider. The loss rate function represents the percentage of loss in the event of default or failure. This represents the risk exposure function, used to characterize the exposure balance or exposure amount in the event of default or failure, where Π(x) represents the expected return term. This includes resource costs and operating costs.

[0038] Specifically, the resource-side feature vector corresponding to the i-th sample... Consumer feature vector The process of constructing a risk estimation model for resource provider default or failure, using the phased product configuration parameter vector x as input, involves establishing a probability function for resource provider default or failure risk. In the process of constructing a risk estimation model for resource party default or failure, the data quality score corresponding to the i-th sample is used as the basis. Weighting of sample loss terms or statistical estimates can reduce the impact of samples with high missing rates, many outliers, or poor feature consistency verification results on the parameter learning of the resource party default or failure risk estimation model.

[0039] And based on the probability function of resource provider default or failure risk Loss rate function With risk exposure function Construct the expected loss function and in the expected loss function Based on this, combined with expected return items and resource costs and operating costs Construct a risk-adjusted return function as the resource-side objective function. The risk-adjusted payoff function serves as a computational input in the multi-objective optimization module, participating in the selection and ranking of candidate solutions.

[0040] Furthermore, risk threshold constraints and business rule constraints include risk thresholds. Risk-adjusted return lower limit threshold The upper and lower bounds of the allowable range for rate structure parameters are respectively and The upper and lower limits of the permissible down payment ratio are respectively and and the optional set of periods ; in, , , , ; Installment Product Configuration Parameter Vector The expression is: , Constructing multi-objective function vectors Multi-objective function vector The expression is: , in, Indicates the period number. , Indicates the down payment percentage. Indicates the rate structure parameters, Indicates the interest subsidy sharing parameter; The interest subsidy sharing parameter s includes the consumer's interest subsidy sharing ratio. Merchant interest subsidy sharing ratio Interest subsidy sharing ratio with resource providers , Risk threshold constraints and parameter range constraints together constitute a set of candidate solution constraints, which are used to limit the search space of candidate parameters, reduce the number of evaluations of candidate solutions that do not meet the constraints, and reduce the consumption of computational resources in the multi-objective solution process.

[0041] Furthermore, S3 includes the following steps: S3-1. Under the constraint set Ω consisting of risk threshold constraint and parameter range constraint, perform constraint pre-screening on candidate solutions and eliminate candidate solutions that do not satisfy the constraint set Ω. S3-2. Calculate the multi-objective function vector for the pre-screened candidate solutions. ; S3-3. Generate a set of non-dominated solutions using Pareto dominance relations, and incrementally maintain the set of non-dominated solutions. For any two sets of candidate solutions... If the following conditions are met: , , but And the solutions that are not dominated by any other solution form a Pareto optimal solution set; S3-4. Under preset computing resource or response delay constraints, output the top K non-dominated solutions of the Pareto optimal solution set or the solution set that satisfies the coverage threshold. Specifically, under the constraints of pre-set computing resource thresholds or response latency thresholds based on available system computing resources or business response latency requirements, the system outputs the top K non-dominated solutions of the Pareto optimal solution set, or outputs a non-dominated solution set that meets a pre-set solution set coverage threshold. This reduces the computational burden of subsequent sorting and filtering processes while ensuring the representativeness of the solution set. The solution set coverage threshold characterizes the degree to which the output solution set covers the Pareto optimal solution set or the target space.

[0042] S3-5. Using a weighted scoring function Ranking or quadratic decision-making on the set of Pareto optimal solutions, the weighted scoring function is used in the consumer objective function. Merchant objective function Objective function of resource side The weighted scoring function weighs the options and determines the final output solution. The expression is: , and , in, Configure parameters for the target weights, which correspond to the consumer objective function. Merchant objective function and resource objective function ; S3-6. Select the weighted scoring function. The output of the scheme with the largest value is the phased product configuration parameter combination, and an audit record is generated simultaneously for this output scheme. This audit record includes the selected scheme. Corresponding objective function value Constraint satisfaction results, model version number, and filtering strategy identifier.

[0043] Furthermore, after the installment product was launched, indicators such as application approval rate, conversion rate, refund rate, delinquency rate, default rate, and risk-adjusted return were collected to form an effectiveness evaluation vector Yt, which was then processed to obtain the new dataset. Based on the new dataset Model parameter sets for three types of target models Update; The update is executed when a preset update trigger condition is met. The update trigger condition includes any of the following: effect evaluation vector. The model update satisfies the following conditions: the offset from the historical baseline exceeds a threshold, the drift of the input feature distribution exceeds a threshold, or the number of new samples reaches a threshold. ,in For learning rate, For loss function, , This indicates the new dataset used for model updates; If the stability metrics and constraint satisfaction rate do not meet the preset standards in the validation set or sliding time window after the update, the model version before the update will be rolled back. Simultaneously, the range or candidate set of values ​​for the number of periods, down payment ratio, fee structure, and interest subsidy allocation parameters are adjusted to reduce parameter oscillations, improve constraint satisfaction rate, and enhance system output stability.

[0044] Specifically, the system collects metrics such as application approval rate, conversion rate, refund rate, delinquency rate, default rate, and risk-adjusted return after the installment product launch to form an effectiveness evaluation vector Yt. Furthermore, it performs time window alignment, latency compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification on newly collected consumer, merchant, and resource provider data after launch, creating a new dataset for model updates. Based on the new dataset The set of model parameters θ for the three target models is updated so that the model parameters undergo controlled iteration with new samples.

[0045] The multi-objective collaborative optimization home decoration consumer installment product multi-party collaborative configuration system adopts any of the above methods, including: a multi-source data acquisition and processing module, used to acquire and process consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, resource provider cost data and post-service performance data, and perform time window alignment, delay compensation, missing value handling, anomaly detection, feature construction, standardization and feature consistency verification to form a modeling dataset; The three-party target modeling module is used to construct consumer choice response models, merchant conversion revenue models, resource provider risk estimation models, and risk-adjusted revenue models, and outputs the results. , , ; The multi-objective optimization module is used to establish a multi-objective optimization model and generate a Pareto optimal solution set under risk thresholds and business rule constraints, wherein the following conditions are met: Furthermore, by using a constraint pre-screening and non-dominated solution incremental maintenance mechanism, a solution set is output under preset computing resource or response delay constraints, in order to reduce the candidate solution search space, reduce the number of invalid evaluations, and improve the system's solution efficiency. The scheme selection and parameter output configuration module is used to select the Pareto optimal scheme set and output the number of periods, down payment ratio, rate structure and interest subsidy allocation configuration parameters, and generate audit records containing constraint satisfaction records and model version information. The monitoring and feedback update module is used to monitor the effectiveness of the campaign and post-service performance, and to update based on preset trigger conditions. Alternatively, an equivalent update rule can be used to update the model, and a version rollback can be performed if the update fails verification. At the same time, the candidate set or value range of configuration parameters can be adjusted to reduce output fluctuations caused by model update failures and improve system stability.

[0046] In summary, this invention constructs a three-party objective model by performing time alignment, cleaning, and consistency verification on relevant data from consumers, merchants, and resource providers. It then performs multi-objective optimization under risk thresholds and business rule constraints, improving solution efficiency through constraint pre-screening and incremental maintenance of the solution set, and outputting configuration parameters within the computing power or latency budget. This invention also generates constraint satisfaction records and model version information to support auditing and backtracking, and performs controlled updates when the deployment monitoring meets the trigger conditions, including post-update verification and rollback mechanisms for failures, thereby improving the efficiency and stability of online configuration.

[0047] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0048] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A multi-objective optimization method for the multi-party collaborative configuration of home decoration installment products, characterized in that, Includes the following steps: S1. Acquire multi-source data: Acquire multi-source data related to installment product configuration. Multi-source data includes data from consumers, merchants, and resource providers. Process the multi-source data to form a modeling dataset. S2. Construct three types of target models: Based on the modeling dataset, construct three types of target models for consumers, merchants, and resource providers respectively. The three types of target models include: consumer choice response model, merchant conversion revenue model, resource provider default or failure risk estimation model, and risk-adjusted revenue model. S3. Solving for the Pareto optimal solution set: Under the constraints of risk threshold, parameter range, and system computing resources or response delay, a multi-objective optimization model with three types of objective models as objective functions is established, and the Pareto optimal solution set is obtained by pre-screening of candidate solutions and incremental maintenance mechanism of non-dominated solutions. S4. Forming phased product configuration parameters: Sort, filter and output the obtained Pareto optimal solution set to form phased product configuration parameters, and simultaneously output the constraint satisfaction record, model version information and filtering strategy identifier corresponding to the configuration parameters. S5. Update installment product configuration parameters: Apply the installment product configuration parameters to product launch and monitor the launch effect and post-service performance; when the monitoring results meet the preset update trigger conditions, perform controlled feedback updates on the three target models and installment product configuration parameters to achieve dynamic iterative optimization of installment product configuration and output stability control.

2. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 1, characterized in that: Multi-source data includes consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, resource provider cost data and post-service performance data, among which: Consumer credit information includes one or more of the following: consumer credit score, debt level, historical overdue records, income, and stability indicators; Consumer payment capacity data includes one or more of the following: consumer disposable cash flow, monthly spending structure, repayment affordability, and payment preferences; Merchant transaction characteristic data includes one or more of the following: average order value distribution, conversion rate, customer complaint rate, refund rate, order cancellation rate, and product category structure. Merchant performance-related data includes one or more of the following: installation performance time, delivery completion rate, delayed performance rate, after-sales processing time, dispute resolution results, refund completion time, and performance exception records. Resource provider cost data includes one or more of the following: resource occupancy costs, channel costs, and operating costs. The performance data of the resource provider after providing services includes one or more of the following: delinquency rate, default rate, recovery rate, loss rate, and migration rate.

3. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 2, characterized in that: Multi-source data processing includes time window alignment, latency compensation, cleaning, missing value handling, anomaly detection, feature construction, standardization, and feature consistency verification, among which: After performing time window alignment and delay compensation on multi-source data, samples are obtained, and a data quality score is generated for each sample. The data quality score is determined by the missing proportion, outlier proportion, and consistency check results. After cleaning, handling missing values, detecting anomalies, constructing features, and standardizing the multi-source data, a dataset for modeling is formed. Dataset The expression is: ， in, This represents the consumer feature vector corresponding to the i-th sample. This represents the merchant feature vector corresponding to the i-th sample. This represents the resource-side feature vector corresponding to the i-th sample. This represents the historical result label or statistical result corresponding to the i-th sample. This represents the data quality score corresponding to the i-th sample. This represents the total number of samples.

4. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 3, characterized in that: Consumer credit information and payment ability data are input into the consumer choice response model construction process to establish a probability or acceptance function for consumers' choice of different installment product configuration parameters. It outputs consumer target indicators associated with the installment product configuration parameters as the consumer objective function. ; In the process of constructing a consumer choice response model, the obtained data quality score is used as a basis. The samples are weighted to reduce the impact of low-quality data on model stability; Choose either probability or acceptability function satisfy: ,and , in, This represents the Sigmoid function. Represents the feature mapping function. This represents the consumer's choice of response model parameters, where T represents the transpose and x represents the installment product configuration parameter vector.

5. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 4, characterized in that: Establish a merchant conversion revenue function based on merchant transaction characteristic data. And introduce penalties for refunds or disputes. Adjusting the returns, we obtain the merchant's objective function. And based on the obtained data quality score Weight the training samples or statistical estimates; Merchant conversion revenue function The expression is: ， Merchant objective function The expression is: ， in, This represents the merchant conversion revenue mapping function. This represents the parameters of the merchant conversion revenue model. Indicates the penalty coefficient. .

6. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 5, characterized in that: Based on resource provider cost data and post-service performance data, a risk probability estimation model and a loss assessment model are established to output the risk probability. and construct the expected loss In the expected loss Based on this, a risk-adjusted return function is constructed as the resource-side objective function. ; Output risk probability The expression is: ； Expected loss The expression is: ； Resource side objective function The expression is: ； in, Represents the feature mapping function. This represents the parameters of the resource provider default or failure risk estimation model. This represents the feature vector of the resource provider. The loss rate function represents the percentage of loss in the event of default or failure. This represents the risk exposure function, used to characterize the exposure balance or exposure amount in the event of default or failure, where Π(x) represents the expected return term. This includes resource costs and operating costs.

7. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 6, characterized in that: Risk threshold constraints and business rule constraints include risk thresholds Risk-adjusted return lower limit threshold The upper and lower bounds of the allowable range for rate structure parameters are respectively and The upper and lower limits of the permissible down payment ratio are respectively and and the optional set of periods ; in, , , , ; Installment Product Configuration Parameter Vector The expression is: ， Constructing multi-objective function vectors Multi-objective function vector The expression is: ， in, Indicates the period number. , Indicates the down payment percentage. Indicates the rate structure parameters, Indicates the interest subsidy sharing parameter; The interest subsidy sharing parameter s includes the consumer's interest subsidy sharing ratio. Merchant interest subsidy sharing ratio Interest subsidy sharing ratio with resource providers , .

8. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 7, characterized in that: S3 includes the following steps: S3-1. Under the constraint set Ω consisting of risk threshold constraint and parameter range constraint, perform constraint pre-screening on candidate solutions and eliminate candidate solutions that do not satisfy the constraint set Ω. S3-2. Calculate the multi-objective function vector for the pre-screened candidate solutions. ; S3-3. Generate a set of non-dominated solutions using Pareto dominance relations, and incrementally maintain the set of non-dominated solutions. For any two sets of candidate solutions... If the following conditions are met: ， ， but And the solutions that are not dominated by any other solution form a Pareto optimal solution set; S3-4. Under preset computing resource or response delay constraints, output the first K non-dominated solutions of the Pareto optimal solution set or the solution set that satisfies the coverage threshold. S3-5. Using a weighted scoring function Sort or make a quadratic decision on the set of Pareto optimal solutions, using a weighted scoring function. The expression is: , and , in, Configure parameters for the target weights, which correspond to the consumer objective function. Merchant objective function and resource objective function ; S3-6. Select the weighted scoring function. The output of the scheme with the largest value is the phased product configuration parameter combination, and an audit record is generated simultaneously for this output scheme. This audit record includes the selected scheme. Corresponding objective function value Constraint satisfaction results, model version number, and filtering strategy identifier.

9. The multi-objective optimization method for multi-party collaborative configuration of home decoration installment products as described in claim 8, characterized in that: The system collects metrics such as application approval rate, conversion rate, refund rate, delinquency rate, default rate, and risk-adjusted return after the launch of installment products to form an effectiveness evaluation vector Yt. This vector is then processed to obtain the new dataset. Based on the new dataset Model parameter sets for three types of target models Update; The update is executed when a preset update trigger condition is met. The update trigger condition includes any of the following: effect evaluation vector. The model update satisfies the following conditions: the offset from the historical baseline exceeds a threshold, the drift of the input feature distribution exceeds a threshold, or the number of new samples reaches a threshold. ,in For learning rate, For loss function, , This indicates the new dataset used for model updates; If the stability metrics and constraint satisfaction rate do not meet the preset standards in the validation set or sliding time window after the update, the model version before the update will be rolled back. Simultaneously, the range or candidate set of values ​​for the number of periods, down payment ratio, fee structure, and interest subsidy allocation parameters are adjusted to reduce parameter oscillations, improve constraint satisfaction rate, and enhance system output stability.

10. A multi-objective collaborative optimization system for home decoration installment payment products, employing the method described in any one of claims 1 to 9, characterized in that: include: The multi-source data acquisition and processing module is used to acquire and process consumer credit information and payment ability data, merchant transaction characteristics and performance-related data, resource provider cost data and post-service performance data, and perform time window alignment, delay compensation, missing value handling, anomaly detection, feature construction, standardization and feature consistency verification to form a modeling dataset. The three-party target modeling module is used to construct consumer choice response models, merchant conversion revenue models, resource provider risk estimation models, and risk-adjusted revenue models, and outputs the results. , , ; The multi-objective optimization module is used to establish a multi-objective optimization model and generate a Pareto optimal solution set under risk thresholds and business rule constraints, wherein the following conditions are met: Furthermore, by using a constraint pre-screening and non-dominated solution incremental maintenance mechanism, a solution set is output under preset computing resource or response delay constraints, in order to reduce the candidate solution search space, reduce the number of invalid evaluations, and improve the system's solution efficiency. The scheme selection and parameter output configuration module is used to select the Pareto optimal scheme set and output the number of periods, down payment ratio, rate structure and interest subsidy allocation configuration parameters, and generate audit records containing constraint satisfaction records and model version information. The monitoring and feedback update module is used to monitor the effectiveness of the campaign and post-service performance, and to update based on preset trigger conditions. Alternatively, an equivalent update rule can be used to update the model, and if the update fails verification, a version rollback can be performed. At the same time, the candidate set or value range of configuration parameters can be adjusted to reduce output fluctuations caused by model update failures and improve system stability.

Images