Rental Bond Management System

Abstract

The present invention relates to a rental receivables management system, and more specifically, to an AI-based rental receivables management system capable of analyzing recovery risk and automatically generating recovery strategies for rental receivables arising from long-term arrears in rental fees of water purifiers, air purifiers, bidets, massage chairs, and home appliances based on status information collected from rental products, such as communication connection status, power application status, usage frequency, location information, and maintenance history, and behavioral data, such as the debtor's payment history, contact response history, and application access history; generating explanatory information regarding the basis for calculating recovery risk using Explainable Artificial Intelligence (XAI); verifying forgery by storing the recovery procedures and recovery history of rental products in a blockchain-based distributed ledger; and simultaneously evaluating the feasibility of re-rental and the profitability of re-rental of the recovered rental products. To this end, the present invention comprises: a receivables information collection unit that collects information on rental receivables generated due to arrears in rental fees of rental products; and debtor information corresponding to the said rental receivables. It includes a receivables management unit that stores and manages rental product information and delinquency information, a receivables recovery processing unit that performs legal action, collection, installment repayment, or receivables recovery procedures depending on the recoverability of the rental receivables, and a recovery history management unit that stores the recovery results and recovery history of the rental receivables.

Description

Technology Field

[0001] The present invention relates to a rental receivables management system, and more specifically, to an AI-based rental receivables management system capable of analyzing recovery risk and automatically generating recovery strategies based on status information collected from rental products, such as communication connection status, power application status, usage frequency, location information, and maintenance history, and behavioral data such as the debtor's payment history, contact response history, and application access history, for rental receivables arising from long-term delinquency of rental fees for water purifiers, air purifiers, bidets, massage chairs, and home appliance rental products; generating explanatory information for the basis of calculating recovery risk using explainable artificial intelligence (XAI); verifying forgery by storing the recovery procedures and recovery history of rental products in a blockchain-based distributed ledger; and evaluating the possibility of re-rental and the profitability of re-rental of the recovered rental products. Background Technology

[0002] Recently, as the rental service market for household appliances such as water purifiers, air purifiers, bidets, massage chairs, mattresses, and garment care machines has expanded, the use of long-term installment payment methods based on rental contracts is increasing, and accordingly, rental receivables resulting from long-term arrears in rental fees are also continuously increasing.

[0003] In particular, since rental services are structured to provide services such as product installation, maintenance, filter replacement, and regular inspection over a long period, unlike general financial receivables, they have the characteristic that condition management and collection management of the physical rental product must be performed simultaneously along with receivables management.

[0005] Conventional debt management systems generally employ a method of determining the likelihood of debt recovery based on a debtor's delinquency information, credit information, and payment history, and then carrying out collection or legal action. Additionally, some prior art has proposed technologies that manage debt recovery procedures by calculating debt risk ratings, managing integrated information on multiple debtors, and storing recovery results in a database.

[0006] However, most of these conventional technologies are designed for general financial claims or credit claims of financial institutions, and have limitations in that they do not reflect the usage status, location status, maintenance status, and product recovery possibility of the actual rental product.

[0007] In particular, conventional rental receivable management methods often determine collection priority based on factors such as whether the debtor responds to contact or the simple period of delinquency, which made it difficult to determine in real time whether the rental product is actually being used, whether the installation location has changed, or whether the power has been cut off for a long period.

[0008] Consequently, there was a problem in which recovery costs increased and recovery efficiency decreased due to the need to perform repeated on-site collection or legal action on receivables that already had a significantly low probability of recovery.

[0010] In addition, since most conventional technologies are structured to store the results of debt recovery or the history of legal action in a general database, there was a problem in that it was difficult to verify whether information was altered or history was falsified during the process of carrying out the recovery procedure. In particular, there was a problem in that the reliability of the recovery procedure could be lowered due to the lack of an objective verification system for rental product installation information, location change information, information on on-site recovery, and product handover information.

[0012] Furthermore, although AI-based risk assessment technology is being applied to various financial fields recently, conventional AI-based receivables management technology often remains at the level of simply calculating risk grades, making it difficult to explain which status information or behavioral data influenced the increase in recovery risk. In particular, it was difficult to intuitively grasp the impact of changes in the status of physical rental products—such as long-term non-use, frequency of location changes, communication disconnection status, and usage reduction rates—on recovery risk, which led to a decrease in the efficiency of strategy formulation by managers.

[0014] In addition, conventional rental receivables management systems had limitations in that they could not comprehensively analyze the re-rental potential or reusability of recovered rental products. In other words, because they failed to consider the years of use, wear condition of parts, maintenance history, and market demand of recovered products, there was a problem where products with a high probability of re-rental were inefficiently discarded, or conversely, unnecessary recovery costs were incurred for products with low reusability value.

[0016] Therefore, there is a need to develop a new type of AI-based rental receivables management system that can precisely analyze the recovery risk and recovery possibility of rental products based on status information collected from rental products and debtor behavior data, verify whether the recovery procedure has been tampered with, and comprehensively evaluate the possibility of re-renting recovered rental products. Prior art literature

[0017] Republic of Korea Registered Patent No. 10-2517832 (Registered on March 30, 2023) Republic of Korea Registered Patent No. 10-1667965 (Registered on October 14, 2016) Republic of Korea Registered Patent No. 10-1112755 (Registered on January 30, 2012) The problem to be solved

[0018] The present invention is proposed to solve the above problems and aims to provide an artificial intelligence-based rental receivables management system capable of precisely analyzing whether a rental product is actually used, the possibility of the debtor going into hiding, the possibility of recovering the rental product, and the priority of recovery based on status information collected from the rental product, such as communication connection status, power supply status, usage frequency, location information, and maintenance history, and behavioral data, such as the debtor's payment history, contact response history, and application access history.

[0020] In addition, the objective of the present invention is to provide a rental receivables management system that generates a standard model of the normal usage pattern of a rental product, analyzes the deviation from collected status information to determine whether there is abnormal behavior such as a long-term non-use state, a state of being out of place, a state of communication blockage, and a state of rapid decrease in usage, and based on this, can more accurately predict the possibility of failure to recover the rental product and the risk of recovery.

[0022] In addition, the objective of the present invention is to provide a rental receivables management system that can intuitively provide the basis for calculating recovery risk to the manager and improve the reliability and efficiency of establishing recovery strategies by generating explanatory information through explainable artificial intelligence (XAI) regarding the impact of changes in the status information of rental products, frequency of location changes, communication blockage status, and delinquency progress status on recovery risk.

[0024] In addition, the objective of the present invention is to provide a rental receivables management system that can improve the reliability and traceability of the recovery history by storing information such as installation information, location change information, visit collection information, recovery completion information, product handover information, and legal action execution information of rental products in a blockchain-based distributed ledger, and verifying whether the rental product recovery procedure has been altered or falsified.

[0026] In addition, the objective of the present invention is to provide a rental receivables management system that can improve asset operation efficiency and recovery efficiency of rental products by analyzing the usage years, wear condition of parts, maintenance history, cleaning or restoration costs, and market demand by model to calculate the feasibility of re-rental and profitability of re-rental, and determining whether to prioritize recovery, re-rental, sell as used, or dispose of by comparing with recovery costs.

[0028] In addition, the objective of the present invention is to provide an artificial intelligence-based rental receivables management system that can reduce the cost of recovering rental receivables and improve the success rate of recovery by comprehensively analyzing recovery risk, recoverability, re-rental possibility grade, and expected recovery path, and automatically generating a recovery strategy including visit recovery, legal action, inducing installment repayment, and collection priority. means of solving the problem

[0029] To achieve the above objective, the present invention comprises: a receivable information collection unit that collects rental receivable information generated due to overdue rental fees of a rental product;

[0030] A receivables management unit that stores and manages debtor information, rental product information, and delinquency information corresponding to the above rental receivables;

[0031] A debt recovery processing unit that performs legal action, collection, installment repayment, or debt recovery procedures depending on the recoverability of the above rental receivables;

[0032] It includes a recovery history management unit that stores the recovery results and recovery history of the above rental receivables.

[0034] In addition, the above rental receivables management system is,

[0035] An IoT status collection unit that collects at least one status information from the above rental product, including the product's communication connection status, power application status, operating time, usage frequency, sensor operation information, installation location information, network connection status, maintenance history, filter replacement history, and product abnormal status;

[0036] An abnormal behavior analysis unit that generates a standard model of the normal usage pattern of the rental product and determines whether there is abnormal behavior by analyzing the rate of change in usage pattern, long-term non-use state, duration of power cutoff, whether there is a deviation from the location, whether communication is cut off, and whether usage has decreased sharply by comparing with the status information;

[0037] An AI risk assessment unit that calculates whether the rental product is actually used, the possibility of the debtor going into hiding, the possibility of recovering the rental product, the estimated recovery period, and the recovery priority based on at least one behavioral data among the above status information, abnormal behavior analysis results, rental fee delinquency information, the debtor's payment history, contact response history, application access history, and location change history;

[0038] An explainable artificial intelligence (XAI) processing unit that analyzes the influence of feature values ​​serving as the basis for calculating the recovery risk calculated by the AI ​​risk assessment unit above, and generates explanatory information on the impact of long-term non-use status, frequency of location change, communication blockage status, usage reduction rate, and delinquency progress status on the recovery risk;

[0039] An autonomous recovery strategy generation unit that automatically generates a recovery strategy including on-site recovery, legal action, induction of installment payments, collection priority, and allocation of a recovery schedule based on the recovery probability, estimated recovery period, and residual value of the rental product calculated by the AI ​​risk assessment unit;

[0040] A re-rental value assessment unit that calculates the reusability of the rental product, the re-rental eligibility grade, and the estimated re-rental revenue by analyzing the years of use, frequency of use, maintenance history, presence of external damage, condition of parts wear, and used transaction and re-rental history by model of the above rental product;

[0041] It is further characterized by including a non-face-to-face recovery support unit that outputs payment guidance information, recovery schedule information, installment repayment proposal information, product recovery request information, or service restriction guidance information to a debtor terminal or rental product according to a recovery strategy generated by the above-mentioned autonomous recovery strategy generation unit.

[0043] At this time, the AI ​​risk assessment unit comprises: a risk analysis module that calculates a recovery risk by analyzing the rate of change in usage patterns, frequency of location movement, duration of long-term non-use, frequency of communication blocking, and the status of delinquency of the rental product as time-series data based on the status information and behavior data; an abnormal behavior risk prediction module that calculates an abnormal behavior score by calculating the deviation between the normal usage pattern reference model of the rental product and the current status information, and predicts the possibility of disappearance and the possibility of recovery failure based on the abnormal behavior score; a recovery path prediction module that calculates the recovery-possible area, visit recovery priority, and expected recovery path of the rental product based on the location information, usage status, and past recovery success history of the rental product; and a strategy decision module that calculates a recovery strategy priority by synthesizing the recovery risk, abnormal behavior score, recovery-possible area, and re-rental eligibility grade.

[0044] The above-mentioned explainable artificial intelligence (XAI) processing unit is,

[0045] A feature importance analysis module that calculates the influence of feature values ​​used in the above-mentioned risk analysis module and abnormal behavior risk prediction module, and generates explanatory information on the contribution of long-term non-use status, frequency of location change, communication blocking status, usage reduction rate, and delinquency progress status to the increase in recovery risk, and

[0046] A risk analysis output module that converts the actual usage status, long-term non-use status, out-of-location status, recovery risk status, and recovery priority of a rental product into visualized risk analysis information based on the above-mentioned descriptive information and outputs it to an administrator terminal, and

[0047] It includes a recovery history verification module that verifies whether the rental product recovery procedure has been tampered with by storing the above-mentioned explanatory information, recovery risk level, recovery power, rental product location change information, visit recovery execution information, recovery completion information, product handover information, and legal action execution information in a blockchain-based distributed ledger, and

[0048] The above-mentioned re-rental value assessment department,

[0049] A re-rental profitability analysis module that calculates re-rental profitability by analyzing the product condition of recovered rental products, parts replacement costs, cleaning or restoration costs, expected re-rental period, and market demand by model, and

[0050] It is characterized by including a recycling decision module that compares the profitability of re-rental and recovery costs to determine whether to prioritize recovery, re-rental, sell as used, or dispose of. Effects of the invention

[0051] The present invention, as described above, has the effect of more accurately determining whether the rental product is actually being used, the possibility of the debtor going into hiding, and the possibility of recovering the rental product by comprehensively analyzing status information such as communication connection status, power supply status, usage frequency, location information, and maintenance history collected from the rental product, and behavioral data such as the debtor's payment history, contact response history, and application access history.

[0053] In addition, the present invention generates a standard model of the normal usage pattern of a rental product and analyzes the deviation from the collected status information to determine whether there is abnormal behavior such as a long-term non-use state, a state of being out of place, a state of communication blockage, and a state of rapid decrease in usage, thereby predicting the possibility of recovery failure in advance and reducing recovery costs by reducing unnecessary visits or legal actions for receivables with a low probability of recovery.

[0055] In addition, the present invention has the effect of enabling a manager to intuitively verify the basis for calculating the recovery risk and improving the reliability of establishing recovery strategies and decision-making efficiency by generating explanatory information on the impact of changes in the condition of rental products, frequency of location changes, communication blockage status, and delinquency progress status on the recovery risk using explainable artificial intelligence (XAI).

[0057] In addition, the present invention stores installation information, location change information, visit collection information, collection completion information, product handover information, and legal action execution information of rental products in a blockchain-based distributed ledger, thereby verifying whether the rental product collection procedure has been tampered with. This improves the reliability and traceability of the collection history and has the effect of being used as objective verification material in the event of a dispute.

[0059] In addition, the present invention analyzes the usage years, wear condition of parts, maintenance history, cleaning or restoration costs, and market demand of a rental product to calculate whether it is possible to re-rent and the profitability of re-rental, and determines whether to re-rent, sell as used, or dispose of it by comparing it with recovery costs, thereby having the effect of improving the asset utilization efficiency of a rental product and increasing the profitability of re-rental.

[0061] In addition, the present invention can efficiently perform recovery procedures and improve the recovery success rate, and has the effect of improving the automation and operational efficiency of rental receivable management by automatically generating a recovery strategy including visit recovery, legal action, inducing installment repayment, and collection priority based on recovery risk, recovery possibility, expected recovery path, and re-rental possibility grade.

[0063] In addition, the present invention can implement a rental asset-specialized recovery management system that is differentiated from conventional receivables management systems centered on general financial receivables by linking and analyzing the status information and recovery procedures of physical rental products, and has the effect of enabling integrated management of the entire lifecycle of rental products. Brief explanation of the drawing

[0064] FIG. 1 is an exemplary diagram illustrating the overall configuration of a rental receivables management system according to the present invention. FIG. 2 is an exemplary diagram illustrating in detail the AI ​​risk assessment unit constituting the present invention. FIG. 3 is an exemplary diagram illustrating in detail an explainable artificial intelligence processing unit constituting the present invention. FIG. 4 is an exemplary diagram illustrating in detail the re-rental value evaluation unit constituting the present invention. FIG. 5 is a flowchart illustrating the sequence of a rental receivable recovery processing process according to the present invention. FIG. 6 is a flowchart illustrating the sequence of a blockchain-based recovery history verification process according to the present invention. FIG. 7 is an example diagram illustrating the state in which the rental receivables management system of the present invention is displayed on the administrator dashboard screen. FIG. 8 is a schematic diagram illustrating the IoT interoperability and system linkage configuration of the present invention. Specific details for implementing the invention

[0065] In addition to the above objectives, other objectives and features of the present invention will become apparent through the description of embodiments with reference to the accompanying drawings.

[0067] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0069] Hereinafter, a preferred embodiment of the rental receivables management system according to the present invention will be examined in detail with reference to the attached drawings.

[0071] First, the rental receivable management system (1) according to the present invention includes a receivable information collection unit (10) that collects rental receivable information generated due to overdue rental fees of rental products, a receivable management unit (20) that stores and manages debtor information, rental product information, and overdue information corresponding to the rental receivable, a receivable recovery processing unit (30) that performs legal action, collection, installment repayment, or receivable recovery procedures according to the possibility of recovering the rental receivable, and a recovery history management unit (40) that stores the recovery results and recovery history of the rental receivable.

[0073] The above-mentioned receivable information collection unit (10) is configured to collect rental receivable information generated due to overdue rental fees of rental products, and can be connected via a wired or wireless communication network to at least one external system among a rental business operator server that manufactures or rents rental products, a financial institution server, an electronic payment server, a customer management server (CRM), an IoT management server, and an external credit information server.

[0075] The above-mentioned debt information collection unit (10) can collect rental contract information, debtor information, rental product information, overdue payment information, payment history information, automatic transfer failure information, contract termination information and recovery request information, etc.

[0076] At this time, the above rental contract information may include the rental start date, rental period, monthly rental fee, agreed conditions, contract status and service provision conditions, etc., and the above debtor information may include the debtor's name, contact information, address, email address, payment method information, credit rating information and identity verification information.

[0077] In addition, the above rental product information may include the product name, product model name, product unique identification number, installation location, product installation date, maintenance history, filter replacement history, product replacement history, and product return history of the rental product. In particular, the above product unique identification number may include at least one of a serial number, MAC address, communication module identification number, or blockchain-based product identification key assigned to each product, thereby enabling the individual tracking and management of status information and return history for each rental product.

[0078] In addition, the above-mentioned debt information collection unit (10) can collect IoT status information collected from rental products in conjunction. The above-mentioned IoT status information may include the power supply status, usage frequency, usage time, communication connection status, network connection status, sensor operation information, location information, maintenance status, error occurrence status, and product abnormality status of the rental products. Accordingly, the above-mentioned debt information collection unit (10) can collect and manage not only simple overdue information but also the actual usage status and installation status of the physical rental products.

[0079] In addition, the debt information collection unit (10) may additionally collect behavioral data of the debtor. The behavioral data may include application access history, login history, location change history, customer center inquiry history, text message response history, phone connection history, whether payment promises have been fulfilled, and the status of installment repayment progress. At this time, the behavioral data may subsequently be used as training data or analysis data to calculate the likelihood of the debtor going into hiding, the likelihood of recovery, and the priority of recovery by the AI ​​risk assessment unit (70).

[0080] In addition, the debt information collection unit (10) can perform data consistency verification, removal of duplicate data, filtering of abnormal data, and encryption of personal information on the collected debt information and status information. For example, the debt information collection unit (10) can determine whether duplicate debts are registered for the same debtor, or classify location information or usage information that falls outside the normal range as abnormal data and manage it separately.

[0081] In addition, the above-mentioned receivable information collection unit (10) can provide the collected receivable information to the receivable management unit (20), the AI ​​risk assessment unit (70), the explainable artificial intelligence (XAI) processing unit (80), and the re-rental value assessment unit (100), and can store history information regarding the time of receivable creation, the time of delinquency occurrence, the time of recovery request, and the time of product recovery completion by linking with a blockchain-based distributed ledger. Accordingly, there is an effect of being able to comprehensively manage lifecycle information from the creation of rental receivables to the completion of recovery.

[0083] The above-mentioned receivables management unit (20) is configured to store and manage rental receivables information, debtor information, rental product information, delinquency information, and recovery progress information collected from the receivables information collection unit (10), and may be configured to include a database server, a cloud storage device, or a distributed storage system.

[0085] The above-mentioned receivables management unit (20) can generate and manage receivables numbers for each rental contract, and can store debtor information, rental product information, rental fee payment status, time of default, default period, recovery progress status, legal action status and whether recovery is completed, etc., corresponding to the receivables number.

[0086] At this time, the above-mentioned receivables management unit (20) can manage lifecycle information from the installation of the rental product to the occurrence of arrears, the performance of recovery, and the decision on whether to re-rent by mutually mapping the unique identification number for each rental product and the receivables protection.

[0087] In addition, the debt management unit (20) can store behavioral data including the debtor's payment history, automatic transfer failure history, contact response history, customer center consultation history, application access history, and installment repayment progress status, and can provide the behavioral data to the AI ​​risk assessment unit (70) and the explainable artificial intelligence (XAI) processing unit (80) to be used for recovery risk analysis and recovery strategy generation.

[0088] In addition, the above-mentioned receivables management unit (20) can store status information such as the communication connection status, power supply status, usage frequency, usage time, location information, maintenance status, and product abnormality status of the rental product collected from the IoT status collection unit (50), and can use the above-mentioned status information as basic data for analyzing whether the rental product is actually used, whether it is not used for a long period, whether its location has changed, and whether it is recoverable.

[0089] In addition, the above-mentioned receivables management unit (20) can classify and manage the receivables status according to the progress of the overdue status of the rental product into one or more of the following: normal status, short-term overdue status, long-term overdue status, recovery in progress status, legal action status, recovery completed status, and re-rental waiting status. For example, if the overdue period continues for longer than a standard period, the above-mentioned receivables management unit (20) can classify the receivables into a long-term overdue status and request the AI ​​risk assessment unit (70) to analyze the recovery risk.

[0090] In addition, the debt management unit (20) is linked with the autonomous recovery strategy generation unit (90) to update the visit recovery schedule, collection priority, installment repayment progress status, legal action execution status, and recovery completion status in real time, and can provide the recovery priority of multiple debts to the manager's terminal by comparing and analyzing them.

[0091] In addition, the above-mentioned receivables management unit (20) can store explanatory information generated from the explainable artificial intelligence (XAI) processing unit (80) and can provide status information and behavioral data that have influenced the calculation of recovery risk so that they can be checked on the manager terminal. Accordingly, the manager can intuitively identify not only the simple risk level but also the cause of the increase in recovery risk.

[0092] In addition, the above-mentioned receivables management unit (20) is linked with the recovery history management unit (40) and the recovery history verification module (83) to store or verify installation information, location change information, visit recovery performance information, product handover information, recovery completion information, and legal action performance information of rental products in a blockchain-based distributed ledger. Accordingly, it is possible to verify whether the recovery procedure of rental receivables has been tampered with, and the reliability and traceability of the recovery history can be improved.

[0093] In addition, the above-mentioned receivables management unit (20) can be linked with the re-rental value evaluation unit (100) to store and manage information on the usage period, wear condition of parts, maintenance history, cleaning or restoration costs, and re-rental profitability of the recovered rental product, and can provide management information to determine whether to re-rent, sell as used, or dispose of based on this.

[0094] Accordingly, the above-mentioned receivables management department (20) has the effect of being able to comprehensively manage the entire lifecycle information from the creation of rental receivables to the occurrence of delinquency, the performance of recovery, the completion of recovery, and the decision to re-rent.

[0096] The above-mentioned debt recovery processing unit (30) is configured to perform a recovery procedure for rental debt based on rental debt information, delinquency information, recovery progress information provided by the above-mentioned debt management unit (20), and recovery risk, recovery possibility, and recovery priority information provided by the AI ​​risk assessment unit (70), and may be configured to include at least one of a visit recovery execution unit, a legal action execution unit, a non-face-to-face collection execution unit, a recovery schedule management unit, and a recovery result processing unit.

[0098] The above-mentioned debt collection processing unit (30) can determine the collection method and collection priority by comprehensively analyzing the overdue period of the rental product, the risk of collection, the possibility of disappearance, whether the location has changed, whether it has not been used for a long time, whether communication has been blocked, and the re-rental eligibility grade. For example, if the long-term non-use status continues and the frequency of location changes increases, the above-mentioned debt collection processing unit (30) may determine that the risk of disappearance is high and adjust the priority of visit collection upward.

[0099] In addition, the debt recovery processing unit (30) can be linked with the autonomous recovery strategy generation unit (90) to automatically apply a recovery strategy including whether to perform on-site recovery, induce installment repayment, take legal action, request product recovery, and restrict services. At this time, the recovery strategy can be dynamically changed based on the debtor's payment history, contact response rate, product usage status, re-rental value, and past recovery success history.

[0100] In addition, the debt recovery processing unit (30) can be linked with the non-face-to-face recovery support unit (110) to output payment guidance information, overdue warning information, installment repayment proposal information, product recovery request or recovery scheduled information to the debtor terminal or rental product.

[0101] For example, the debt recovery processing unit (30) may send a notification of scheduled recovery to the debtor terminal or output a recovery request guide through the display or voice output means of the rental product when the delinquency period continues for longer than the standard period.

[0102] In addition, the above-mentioned receivable collection processing unit (30) can automatically assign a movement path and collection schedule for a visit collection performer based on location information, usage status, and collection area information of the rental product. Accordingly, the collection path for multiple rental products can be optimized to reduce visit collection costs and collection time.

[0103] Additionally, the debt recovery processing unit (30) can perform at least one legal action among an application for a payment order, a request for seizure of claims, a lawsuit for return, a request for compulsory execution, and a claim transfer procedure through the legal action execution unit. At this time, the debt recovery processing unit (30) can determine whether to perform legal action based on the recovery possibility and estimated recovery period information provided by the AI ​​risk assessment unit (70).

[0104] In addition, the above-mentioned receivables recovery processing unit (30) can determine whether the recovered rental product is reusable, re-rental eligible grade, whether it can be sold as used, or whether it is disposed of by linking with the re-rental value evaluation unit (100) after the recovery of the rental product is completed, and if the profitability of re-rental is low compared to the recovery cost, the recovery procedure can be restricted or the recovery priority can be adjusted.

[0105] In addition, the debt collection processing unit (30) can transmit information on the execution of a visit for collection, information on the completion of collection, information on the handover of a product, information on the execution of legal action, and information on the collection result to the collection history management unit (40) and the collection history verification module (83), and the information can be stored in a blockchain-based distributed ledger to verify whether the collection procedure has been tampered with. For example, the debt collection processing unit (30) can generate objective verification data regarding whether the collection has been completed by storing location information, captured images, electronic signature information, and product acceptance confirmation information collected from the terminal of the collection performer in a blockchain-based distributed ledger.

[0106] In addition, the debt recovery processing unit (30) may provide the AI ​​risk assessment unit (70) with the success or failure of recovery based on the results of the recovery, the time required for recovery, the debtor response rate, the success or failure of installment repayment, and the occurrence of complaints, and the information may be used as training data for a recovery strategy thereafter.

[0107] Accordingly, the above-mentioned receivable recovery processing unit (30) can dynamically execute a recovery strategy based on the status information of the rental product, the debtor's behavior data, and the re-rental value information, and has the effect of improving the efficiency of rental receivable recovery and reducing recovery costs.

[0109] The above recovery history management unit (40) is configured to store and manage the history of the procedures for the on-site recovery, non-face-to-face collection, legal action, product handover, and recovery completion of rental products performed by the debt recovery processing unit (30), and may include a database server, a cloud storage device, a blockchain-linked server, or a distributed storage system.

[0111] The above recovery history management unit (40) can store recovery lifecycle information such as the time of the recovery request, the time of the visit to recover the product, the time of the recovery completion, the time of the product handover, the time of legal action, and the time of re-rental processing, so as to correspond to a unique identification number and a debt number for each rental product. Accordingly, the entire recovery procedure from the occurrence of overdue rental products to the completion of recovery and the decision on whether to re-rent the product can be tracked and managed.

[0112] In addition, the above-mentioned recovery history management unit (40) can store location information, movement path information, visit time information, captured images, video information, electronic signature information, and product acceptance confirmation information collected from the recovery performer's terminal. For example, the above-mentioned recovery history management unit (40) can verify whether an actual visit recovery was performed by storing a rental product image taken by the recovery performer at the recovery site and GPS-based location information in conjunction.

[0113] In addition, the above recovery history management unit (40) can classify and manage the recovery status of the rental product into one or more of the following: recovery request status, scheduled visit status, visit performed status, recovery in progress status, recovery completed status, handover completed status, and re-rental waiting status. At this time, the recovery status can be updated in real time by linking with the receivables management unit (20) and the autonomous recovery strategy generation unit (90).

[0114] In addition, the above recovery history management unit (40) can store information on changes in the location of rental products, information on changes in power status, information on changes in communication connection status, and information on changes in usage status in the form of time-series data, and the above information can be used as learning data or analysis data for recovery risk analysis and abnormal behavior analysis in the AI ​​risk assessment unit (70) and the explainable artificial intelligence (XAI) processing unit (80).

[0115] In addition, the recovery history management unit (40) can store the history of legal action execution and can manage the status of legal action progress, including whether a payment order has been applied for, whether a seizure of claims has been carried out, whether a lawsuit for return has been carried out, whether compulsory execution has been carried out, and whether claims have been assigned. Accordingly, the status of the recovery procedure progress for each claim can be managed in an integrated manner.

[0116] In addition, the recovery history management unit (40) can store the product condition of the recovered rental product, whether there is external damage, the condition of parts wear, whether maintenance is performed, whether cleaning is performed, and whether restoration is performed, and the information can be used as basic data to calculate whether re-rental is possible and re-rental profitability in the re-rental value evaluation unit (100).

[0117] In addition, the above-mentioned recall history management unit (40) can be linked with the recall history verification module (83) to store recall request information, visit recall execution information, product handover information, recall completion information, and legal action execution information in a blockchain-based distributed ledger. At this time, the above-mentioned recall history management unit (40) can verify whether data is changed or tampered with after the recall procedure is performed by generating a hash value for each recall history information and recording it in the blockchain-based distributed ledger.

[0118] In addition, the above recovery history management unit (40) can generate statistical data regarding the success or failure of recovery, the time required for recovery, the debtor response rate, the success or failure of installment repayment, the occurrence of complaints, and the success or failure of re-rental, and the above statistical data can be used as training data for optimizing the recovery strategy.

[0119] In addition, the above recovery history management unit (40) is linked with an administrator terminal and can output the recovery progress status, recovery execution history, legal action progress status, and re-rental processing status for each rental product in the form of a visualized management screen. Accordingly, the administrator can intuitively check the recovery progress status and recovery results for multiple rental receivables.

[0120] Accordingly, the above recovery history management unit (40) can comprehensively manage history information regarding the recovery procedure of rental products, improve the reliability and traceability of the recovery procedure, and have the effect of optimizing the recovery strategy and improving the efficiency of re-rental asset management.

[0122] Meanwhile, the rental receivables management system of the present invention comprises: an IoT status collection unit (50) that collects at least one status information from the rental product, including the product's communication connection status, power application status, operating time, usage frequency, sensor operation information, installation location information, network connection status, maintenance history, filter replacement history, and product abnormal status; an abnormal behavior analysis unit (60) that generates a standard model for the normal usage pattern of the rental product and determines whether there is abnormal behavior by analyzing the usage pattern change rate, long-term non-use status, power cut-off duration, location deviation, communication cut-off status, and rapid decrease in usage by comparing with the status information; an AI risk assessment unit (70) that calculates whether the rental product is actually used, the possibility of the debtor going into hiding, the possibility of recovering the rental product, the estimated recovery period, and the recovery priority based on at least one behavior data among the status information, abnormal behavior analysis results, rental fee delinquency information, debtor's payment history, contact response history, application access history, and location change history; and the recovery calculated by the AI ​​risk assessment unit (70). An explainable artificial intelligence (XAI) processing unit (80) that analyzes the influence of feature values ​​serving as the basis for calculating risk and generates explanatory information on the impact of long-term non-use status, frequency of location change, communication blockage status, usage reduction rate, and delinquency progress status on recovery risk; an autonomous recovery strategy generation unit (90) that automatically generates a recovery strategy including visit recovery, legal action, induction of installment repayment, collection priority, and allocation of recovery schedule based on the recovery probability, estimated recovery period, and residual value of the rental product calculated by the AI ​​risk assessment unit (70); and a re-rental value assessment unit (100) that analyzes the usage years, usage frequency, maintenance history, presence of exterior damage, condition of parts wear, used transaction history by model, and re-rental history of the rental product to calculate whether the rental product is reusable, the re-rental eligibility grade, and the estimated re-rental profit.It further includes a non-face-to-face recovery support unit (110) that outputs payment guidance information, recovery schedule information, installment repayment proposal information, product recovery request information, or service restriction guidance information to a debtor terminal or rental product according to the recovery strategy generated by the above-mentioned autonomous recovery strategy generation unit (90).

[0124] The IoT status collection unit (50) collects at least one of the following status information from the rental product: the communication connection status, power supply status, operating time, usage frequency, sensor operation information, installation location information, network connection status, maintenance history, filter replacement history, and product abnormality status. As a configuration for collecting status information from the rental product in real time or periodically, it may be configured to include at least one of a communication module, a status detection sensor, a location tracking sensor, a network interface, and a data collection server.

[0126] The above IoT status collection unit (50) can be connected via wired and wireless communication networks with an IoT communication module equipped in a water purifier, air purifier, bidet, massage chair, clothing care machine, mattress and other home appliance rental products, and can collect status information using at least one communication method among Wi-Fi, Bluetooth, LTE, 5G, Zigbee, LoRa, NFC or wired network methods.

[0127] The above IoT status collection unit (50) can collect at least one status information among the power supply status of the rental product, power On / Off duration, usage frequency, usage time, operating mode, sensor operation status, communication connection status, network connection status, filter replacement status, whether maintenance is performed, whether an error occurs, and product abnormality status.

[0128] For example, the IoT status collection unit (50) can collect the filter replacement cycle of a water purifier, the fan operation time of an air purifier, the number of times a bidet is used, or the operating time of a massage chair.

[0129] Additionally, the IoT status collection unit (50) can collect installation location information, whether the location has been moved, and the frequency of location changes of the rental product by using a GPS sensor, a motion detection sensor, or a network-based location tracking function provided in the rental product. Accordingly, the IoT status collection unit (50) can determine whether the rental product has been moved from its initial installation location or whether it has been in a specific location for a long period of time.

[0130] In addition, the IoT status collection unit (50) can store and manage usage pattern data of the rental product in the form of time-series data. The usage pattern data may include usage frequency by time period, daily usage change, power application cycle, communication connection frequency and maintenance performance cycle, and the data may be used as basic data for generating a normal usage pattern reference model or determining whether there is abnormal behavior in the abnormal behavior analysis unit (60).

[0131] Additionally, the IoT status collection unit (50) can determine that a rental product is in a state of long-term non-use, a state of power being cut off for a long time, a state of abnormal location movement, a state of communication being cut off, or a state of rapid decrease in usage is an abnormal state, and can provide the abnormal state information to the accounts receivable management unit (20), the AI ​​risk assessment unit (70), and the explainable artificial intelligence (XAI) processing unit (80). For example, the IoT status collection unit (50) can determine that there is a high possibility of a disappearance if product usage is not detected for a certain period of time or if the communication connection is cut off.

[0132] In addition, the IoT status collection unit (50) can perform data consistency verification and abnormal data filtering for the collected status information. For example, the IoT status collection unit (50) can remove or correct location information that falls outside the normal range or duplicate collected status information, and can separately classify and manage abnormal data caused by sensor errors.

[0133] In addition, the IoT status collection unit (50) can store unique identification numbers and status information for each rental product by mutual mapping, and can manage rental receivable information and status information in an integrated manner by linking with the receivables management unit (20). Accordingly, the correlation between the change in the status of the rental product and the progress of delinquency can be analyzed.

[0134] In addition, the IoT state collection unit (50) can store the collected state information in a blockchain-based distributed ledger by linking it with the recovery history management unit (40) and the recovery history verification module (83). At this time, by generating a hash value for the state information and recording it in the blockchain-based distributed ledger, it is possible to verify whether the state information has been changed or tampered with.

[0135] Additionally, the IoT status collection unit (50) may provide the collected status information to the AI ​​risk assessment unit (70), the abnormal behavior analysis unit (60), the explainable artificial intelligence (XAI) processing unit (80), and the re-rental value assessment unit (100), and the status information may be used to calculate recovery risk, analyze abnormal behavior, generate recovery strategies, evaluate re-rental profitability, and determine recovery priorities.

[0136] Accordingly, the IoT status collection unit (50) can track and manage the actual usage status and location status of the rental product in real time, and has the effect of precisely analyzing the possibility of recovery and the risk of disappearance of the rental product.

[0138] The above abnormal behavior analysis unit (60) is configured to analyze the status information of the rental product provided by the IoT status collection unit (50) to detect abnormal usage status, location status, or communication status of the rental product, and to determine the possibility of the debtor going into hiding and the risk of recovery of the rental product based thereon, and may be configured to include at least one of a normal pattern generation module, a status change analysis module, an abnormal behavior detection module, and a risk indicator judgment module.

[0140] The above abnormal behavior analysis unit (60) can generate a normal usage pattern reference model based on the usage frequency, usage time, power supply status, communication connection status, maintenance history, location information, and sensor operation information of the rental product. At this time, the normal usage pattern reference model can be generated by classifying it by product type, installation environment, user type, or usage period.

[0141] For example, in the case of a water purifier, a normal usage pattern can be generated based on daily purified water usage, water dispensing frequency, filter replacement cycle, and communication connection cycle, and in the case of an air purifier, a normal usage pattern can be generated based on daily operating time, airflow setting frequency, and filter usage.

[0143] In addition, the abnormal behavior analysis unit (60) can calculate the rate of change in usage patterns by analyzing the deviation between the currently collected status information and the normal usage pattern reference model. At this time, the rate of change in usage patterns may include at least one of the following: a rate of decrease in usage, a rate of decrease in usage frequency, a rate of change in power application frequency, a rate of change in communication connection frequency, and a rate of change in maintenance performance cycle.

[0144] In addition, the above abnormal behavior analysis unit (60) may determine that the product is in a long-term non-use state if product usage is not detected for a certain period of time or longer, and may determine that the power is cut off for a long-term power cut-off state if the power is cut off for a reference time or longer. In addition, it may determine that the communication is cut off state if the transmission of status information through the communication module is interrupted.

[0145] In addition, the above abnormal behavior analysis unit (60) can determine whether the location has been moved by comparing the installation location of the rental product with the current location, and can detect an abnormal location movement state by analyzing the number of location changes, the distance traveled, and the movement cycle. For example, if the location is repeatedly moved to a different area from the initial installation location or if location information is collected from multiple areas within a short period of time, it can be determined to be an abnormal location movement state.

[0146] Additionally, the above abnormal behavior analysis unit (60) can calculate an Anomaly Score when at least one of a long-term non-use state, a long-term power cut-off state, a communication cut-off state, an abnormal location change state, and a sudden decrease in usage occurs. At this time, the weight for each state can be set based on the product type, the period of delinquency, the history of past successful recovery, and the history of failed recovery.

[0147] Additionally, the abnormal behavior analysis unit (60) may increase the abnormal behavior score when multiple abnormal behaviors occur simultaneously, and may determine a latent risk state or an increased recovery risk state when the abnormal behavior score is greater than or equal to a threshold value. For example, if a long-term non-use state and a communication blocking state occur simultaneously, a higher risk grade than a simple long-term non-use state may be assigned.

[0148] In addition, the abnormal behavior analysis unit (60) can store the time of occurrence of abnormal behavior, the type of abnormal behavior, the duration of abnormal behavior, and the abnormal behavior score in the form of time series data, and the information can be provided to the AI ​​risk assessment unit (70) and used as analysis data for calculating the recovery risk.

[0149] In addition, the above abnormal behavior analysis unit (60) can be linked with an explainable artificial intelligence (XAI) processing unit (80) to generate the cause of abnormal behavior, the basis for calculating the abnormal behavior score, and factors increasing the risk of recovery in the form of explanatory information. For example, explanatory information such as "85% decrease in usage over the last 30 days," "moved 120 km relative to the installation location," and "no communication connection for 14 days" can be generated.

[0150] In addition, the above abnormal behavior analysis unit (60) is linked with the recovery history management unit (40) to utilize past recovery success cases and recovery failure cases as learning data, and can continuously update the normal usage pattern standard model and abnormal behavior judgment criteria.

[0151] Accordingly, the above abnormal behavior analysis unit (60) can predict in advance the possibility of the debtor going into hiding and the risk of recovery of the rental product based on the status information of the rental product, and has the effect of improving the accuracy of the recovery strategy and the success rate of recovery.

[0153] The above AI risk assessment unit (70) is configured to calculate the recovery risk of a rental product, the possibility of the debtor going into hiding, the possibility of the rental product being recovered, the estimated recovery period, and the recovery priority by analyzing status information, behavior data, delinquency information, and recovery history information provided by the IoT status collection unit (50), abnormal behavior analysis unit (60), debt management unit (20), and recovery history management unit (40), and may include at least one of a risk analysis module, a behavior prediction module, a recovery success prediction module, a recovery path prediction module, and a strategy decision module.

[0155] The above AI risk assessment unit (70) can analyze the debtor's willingness to repay and the possibility of cooperation in recovery by collecting behavioral data including the debtor's payment history, delinquency period, number of delinquencies, automatic transfer failure history, contact response rate, customer center consultation history, application access history, and whether installment repayment was performed.

[0156] In addition, the AI ​​risk assessment unit (70) can determine whether the rental product is actually in use and its current operating status by analyzing status information collected from the IoT status collection unit (50), including power supply status, usage frequency, usage time, location information, communication connection status, maintenance history, and product abnormality status.

[0157] In addition, the AI ​​risk assessment unit (70) can calculate the possibility of going into hiding by analyzing the anomaly score provided by the anomaly analysis unit (60), long-term non-use status, communication blockage status, location deviation status, and usage decrease status. For example, a high risk of going into hiding can be assigned if product use is suspended for a long period, communication connection is blocked, and location change occurs.

[0158] In addition, the AI ​​risk assessment unit (70) can calculate the probability of successful recovery by utilizing past cases of successful recovery and cases of failed recovery as training data. At this time, the training data may include the delinquency period, product type, product condition, debtor behavior pattern, location movement pattern, results of recovery execution, and results of legal action execution.

[0159] In addition, the AI ​​risk assessment unit (70) can calculate the recovery risk using at least one artificial intelligence model among machine learning, deep neural network, random forest, gradient boosting, recurrent neural network, or time series prediction model.

[0160] In addition, the AI ​​risk assessment unit (70) can calculate the trend of change in recovery risk by analyzing changes in usage patterns, frequency of location movement, frequency of communication connection, and the status of delinquency over time in the form of time-series data. Accordingly, it is possible to assess risk based on long-term behavioral pattern changes rather than simple status information at a specific point in time.

[0161] In addition, the AI ​​risk assessment unit (70) can calculate the recovery possible areas and expected recovery routes by analyzing the location information of rental products, past recovery locations, recovery success rates by region, and visit recovery history. For example, if the recovery success rate of the same product group is high in a specific area, a priority recovery strategy can be assigned to that area.

[0162] In addition, the AI ​​risk assessment unit (70) can adjust the recovery priority by analyzing the re-rental possibility grade, expected re-rental profit, and product residual value information provided by the re-rental value assessment unit (100). For example, among multiple rental products with similar recovery potential, a priority recovery strategy can be generated for a product with a high re-rental value.

[0163] In addition, the AI ​​risk assessment unit (70) can calculate a Recovery Priority Score by comprehensively analyzing the recovery risk, the possibility of disappearance, the probability of successful recovery, the estimated cost of recovery, the estimated time required for recovery, and the re-rental eligibility grade. At this time, the Recovery Priority Score can be used as reference information to determine whether to perform on-site recovery, whether to take legal action, whether to induce installment repayment, and whether to perform non-face-to-face collection.

[0164] In addition, the AI ​​risk assessment unit (70) can continuously learn whether recovery is completed, the time required for recovery, whether installment repayment is successful, the debtor response rate, whether complaints occur, and whether re-rental is successful, thereby updating the artificial intelligence model and improving the accuracy of the recovery strategy.

[0165] Additionally, the AI ​​risk assessment unit (70) may provide the calculated recovery risk, possibility of disappearance, possibility of recovery, estimated recovery period, and recovery priority to the explainable artificial intelligence (XAI) processing unit (80), and the XAI processing unit (80) may generate the basis for calculating each evaluation result as explanation information and provide it to the administrator terminal.

[0166] Accordingly, the AI ​​risk assessment unit (70) can accurately predict the probability of successful recovery by comprehensively analyzing the status information of the actual rental product and the behavioral data of the debtor, rather than just simple delinquency information, and has the effect of improving the efficiency of the recovery strategy and reducing the cost of recovering rental receivables.

[0168] Meanwhile, the AI ​​risk assessment unit (70) may include a risk analysis module (71) that calculates the recovery risk by analyzing the rate of change in the rental product's usage pattern, frequency of location movement, duration of long-term non-use, frequency of communication blocking, and the status of delinquency as time series data based on the status information and behavior data; an abnormal behavior risk prediction module (72) that calculates an abnormal behavior score by calculating the deviation between the rental product's normal usage pattern reference model and current status information, and predicts the possibility of disappearance and the possibility of recovery failure based on the abnormal behavior score; a recovery path prediction module (73) that calculates the recovery possible area, visit recovery priority, and expected recovery path of the rental product based on the rental product's location information, usage status, and past recovery success history; and a strategy decision module (74) that calculates the recovery strategy priority by synthesizing the recovery risk, abnormal behavior score, recovery possible area, and re-rental possibility grade.

[0170] The above risk analysis module (71) is configured to calculate the recovery risk score of a rental product by analyzing status information provided from the IoT status collection unit (50), abnormal behavior score provided from the abnormal behavior analysis unit (60), and delinquency information and behavior data provided from the debt management unit (20), and may be configured to include a data preprocessing unit (711), a data feature value extraction unit (712), a risk index calculation unit (713), and a risk grade determination unit (714).

[0172] The above data preprocessing unit (711) can normalize power supply status, usage frequency, usage time, location information, communication connection status, maintenance history, product malfunction status, and debtor's payment history, contact response history, application access history, location change history, and installment repayment history collected from rental products, and correct missing values ​​to convert them into an analyzable dataset.

[0173] The above data feature value extraction unit (712) can extract a feature value that affects the calculation of recovery risk from the above dataset. At this time, the feature value may include at least one of the following: duration of long-term non-use, usage reduction rate, duration of power cutoff, duration of communication cutoff, frequency of location change, distance of location change, period of delinquency, amount of delinquency, contact response rate, installment repayment fulfillment rate, application access frequency, number of product defect occurrences, and history of past recovery success.

[0174] In addition, the data feature value extraction unit (712) can calculate the rate of change in usage patterns by analyzing the trend of change in the state information of the rental product, and can generate abnormal behavior related feature values ​​by calculating the deviation of the current state information compared to the normal usage pattern reference model.

[0175] In addition, the above risk index calculation unit (713) can calculate at least one detailed risk index among the risk index of concealment, risk index of recovery failure, risk index of product concealment, risk index of recovery delay, and risk index of need for legal action by applying weights to the extracted feature values.

[0176] For example, the above risk index calculation unit (713) can increase the risk of concealment when a long-term non-use state and a communication blockage state occur simultaneously, and can increase the risk of product concealment when the frequency of location movement and the distance of movement are greater than or equal to a reference value.

[0177] In addition, the above-mentioned risk index calculation unit (713) can analyze the temporal change pattern of status information using a time series analysis technique. At this time, the risk index can be dynamically calculated by reflecting the trend of decreasing usage, the trend of location movement, the trend of changes in communication connections, and the trend of delinquency progress over a recent period.

[0178] In addition, the risk index calculation unit (713) can predict the recovery risk using at least one of a machine learning model, a deep neural network model, a random forest model, a gradient boosting model, a time series prediction model, or an ensemble model. At this time, the model can be trained using past recovery success cases, recovery failure cases, results of legal action, and results of re-rental as training data.

[0179] In addition, the above-mentioned risk grade determination unit (714) can classify rental receivables into a low-risk group, a caution group, a high-risk group, and an emergency recovery group based on the calculated recovery risk level. For example, if the recovery risk level is less than the first threshold value, it can be classified as a low-risk group, and if it is greater than or equal to the second threshold value, it can be classified as an emergency recovery group.

[0180] In addition, the risk grade determination unit (714) can calculate the recovery priority by considering the recovery risk and the re-rental possibility grade provided by the re-rental value evaluation unit (100). For example, a rental product with a high recovery risk but a very high re-rental value can be selected as a priority recovery target.

[0181] Additionally, the above-mentioned risk analysis module (71) can provide the calculated recovery risk, detailed risk index, risk grade, and feature value information used in calculating the risk to the explainable artificial intelligence (XAI) processing unit (80), and the above-mentioned XAI processing unit (80) can analyze the contribution of each feature value and generate the basis for calculating the risk as explanation information.

[0182] In addition, the above risk analysis module (71) can continuously collect actual recovery results, recovery time, legal action results, whether product recovery was successful and whether re-rental was successful to update the learning data, thereby improving the accuracy of risk prediction.

[0183] Accordingly, the above-mentioned risk analysis module (71), unlike conventional debt evaluation methods that use only simple overdue amounts or credit information, can accurately calculate the actual recovery risk by comprehensively analyzing real-time status information of rental products, behavioral data of debtors, and past recovery history, and has the effect of improving the accuracy of recovery strategies and the success rate of recovery.

[0185] The above abnormal behavior risk prediction module (72) is configured to predict the possibility of the debtor going into hiding, the possibility of the rental product being hidden, the possibility of recovery failure, and the necessity of forced recovery by analyzing the abnormal behavior score (Anomaly Score), status information, and behavior data provided by the abnormal behavior analysis unit (60), and may be configured to include an abnormal behavior pattern analysis unit (721), a risk prediction unit (722), a risk level update unit (723), and a prediction result generation unit (724).

[0187] The above abnormal behavior pattern analysis unit (721) can analyze abnormal behavior data in the form of time series data, including long-term non-use state, power cut-off state, communication cut-off state, location deviation state, rapid decrease in usage state and debtor's out of contact state when renting a product.

[0188] In addition, the above abnormal behavior pattern analysis unit (721) can analyze not only whether a single abnormal behavior occurs, but also the order in which multiple abnormal behaviors occur, the frequency of occurrence, the interval between occurrences, and the duration. For example, a pattern in which communication is blocked after a decrease in usage and then a change in location occurs can be determined as a pattern with a high risk of disappearance.

[0189] The above abnormal behavior pattern analysis unit (721) can analyze abnormal behavior data, including long-term non-use of rental products, power cut-off, communication cut-off, location deviation, rapid decrease in usage, and debtor's out of contact, in the form of time series data.

[0190] In addition, the above abnormal behavior pattern analysis unit (721) can analyze not only whether a single abnormal behavior occurs, but also the order in which multiple abnormal behaviors occur, the frequency of occurrence, the interval between occurrences, and the duration. For example, a pattern in which communication is blocked after a decrease in usage and then a change in location occurs can be determined as a pattern with a high risk of disappearance.

[0191] In addition, the above abnormal behavior pattern analysis unit (721) can calculate the similarity by comparing and analyzing the abnormal behavior patterns included in past recovery success cases and recovery failure cases with the current abnormal behavior pattern. At this time, if the similarity is greater than or equal to a threshold value, it can be determined that the risk state is similar to the past recovery failure pattern.

[0192] In addition, the above-mentioned risk prediction unit (722) can calculate the possibility of going into hiding by using abnormal behavior score, frequency of abnormal behavior occurrence, distance traveled, duration of communication blockage, period of delinquency, contact response rate, and installment repayment fulfillment rate as input data.

[0193] For example, if a state of prolonged non-use persists, a communication blockage is maintained for a certain period or longer, and the response rate to contact decreases sharply, it can increase the likelihood of going into hiding.

[0194] In addition, the risk prediction unit (722) can calculate the possibility of product concealment by analyzing changes in location information and movement paths of rental products. For example, the risk of product concealment can be increased if the product is repeatedly moved to a different area from the initial installation area or if its location is confirmed in multiple areas within a short period of time.

[0195] In addition, the above risk prediction unit (722) can calculate the probability of recovery failure based on abnormal behavior patterns and past recovery result data. At this time, if an abnormal behavior pattern that appeared commonly in past recovery failure cases is also confirmed in the current rental product, the probability of recovery failure can be adjusted upward.

[0196] In addition, the risk prediction unit (722) can predict the likelihood of risk occurring within a certain period in the future by using at least one of a machine learning model, a deep neural network model, a recurrent neural network (RNN), a long short-term memory neural network (LSTM), a time series prediction model, or an ensemble model.

[0197] Additionally, the risk update unit (723) can update the prediction results in real-time or periodically when new status information is received from the IoT status collection unit (50) or when the debtor's behavioral data changes. For example, if a debtor who has not been in contact for a long time applies for installment repayment, the likelihood of disappearance can be lowered.

[0198] In addition, the above-mentioned risk update unit (723) can continuously improve the risk prediction model by reflecting the results of the visit recovery, the results of legal action, whether the recovery was successful, and the results of the re-rental as learning data.

[0199] In addition, the prediction result generation unit (724) can generate a score for the possibility of concealment, a score for the possibility of failure to recover, a score for the possibility of product concealment, and a score for the necessity of forced recovery, and based on this, can classify into a high-risk group, a caution group, or a normal group.

[0200] Additionally, the prediction result generation unit (724) can provide the calculated prediction result to the strategy decision module (74) and the autonomous recovery strategy generation unit (90), and the autonomous recovery strategy generation unit (90) can increase the priority of visit recovery or prioritize legal action procedures when there is a high probability of disappearance.

[0201] Additionally, the above abnormal behavior risk prediction module (72) can provide the feature value and impact, which serve as the basis for calculating the possibility of disappearance, the possibility of recovery failure, and the possibility of product concealment, to the explainable artificial intelligence (XAI) processing unit (80), and the XAI processing unit (80) can generate explanatory information about the cause of the risk occurrence and provide it to the administrator terminal.

[0202] Accordingly, the above abnormal behavior risk prediction module (72) can go beyond simply detecting abnormal behavior in the current state and can predict in advance the risk of potential future disappearance, product concealment, and recovery failure, and has the effect of improving the accuracy of the recovery strategy and the recovery success rate.

[0204] The above recovery path prediction module (73) is configured to predict the recovery area, expected recovery location, priority of visit recovery, and expected recovery path of the rental product by analyzing the location information, status information, behavioral data of the debtor, and past recovery history of the rental product, and may be configured to include a location information analysis unit (731), a movement pattern analysis unit (732), a recovery area calculation unit (733), a recovery path generation unit (734), and a path optimization unit (735).

[0206] The above location information analysis unit (731) can analyze the current location and movement status of the rental product by collecting GPS location information, network-based location information, installation location information, recent connection location information and location change history collected from the IoT status collection unit (50).

[0207] In addition, the location information analysis unit (731) can calculate the distance traveled, the number of moves, and the frequency of location changes by comparing the initial installation location and the current location, and can determine whether the movement is normal or abnormal based on this.

[0208] In addition, the movement pattern analysis unit (732) can analyze the location movement path, movement cycle, movement time period, and stay time of the rental product in the form of time series data. For example, if it is repeatedly located in a specific area, that area can be determined as a major stay area.

[0209] In addition, the above movement pattern analysis unit (732) can estimate the debtor's actual activity area by analyzing the debtor's application access location, payment attempt location, customer center access location, communication connection location, and past residence information in conjunction.

[0210] In addition, the movement pattern analysis unit (732) can learn location pattern data regarding past recovery success cases and recovery failure cases to calculate the influence of a specific location pattern on recovery success.

[0211] In addition, the above-mentioned recovery area calculation unit (733) can calculate candidate areas with a high probability of recovery by comprehensively analyzing the current location of the rental product, the recent place of stay, the debtor's activity area, and past recovery success data.

[0212] For example, if a rental product has been repeatedly connected to a specific area for the past 30 days, that area can be set as a priority recovery area, and if the debtor's mobile application access location is repeatedly confirmed in the same area, the possibility of recovery can be evaluated as even higher.

[0213] In addition, the above-mentioned recoveryable area calculation unit (733) can calculate the probability of recovery success for a plurality of candidate areas and can determine the recovery priority areas in order of the highest probability of recovery success.

[0214] In addition, the above recovery path generation unit (734) can generate a recovery path by considering recovery possible areas, rental product location information, quantity of items to be recovered, current location of the recovery performer, and recovery priority.

[0215] In addition, the above recovery path generation unit (734) can minimize travel distance by generating an integrated recovery path when multiple rental products are located in the same area or adjacent areas.

[0216] In addition, the above-mentioned route optimization unit (735) can calculate the optimal recovery order by considering traffic information, estimated travel time, recovery success probability, the debtor's activity time and possible visit time.

[0217] For example, if a debtor's product usage pattern during specific time periods is identified, that time period can be set as the priority visit time, and the route can be reconfigured to prioritize visiting areas with a high probability of successful recovery.

[0218] In addition, the path optimization unit (735) can dynamically change the recovery path when real-time location information or new status information is received. For example, a new recovery path can be created when the location of the rental product changes or when the debtor requests a change in the recovery schedule.

[0219] Additionally, the recovery path prediction module (73) can provide the expected recovery location, recovery possible area, recovery success probability, and optimal recovery path to the autonomous recovery strategy generation unit (90) and the debt recovery processing unit (30), and the debt recovery processing unit (30) can determine whether to perform a visit recovery and the recovery schedule based on this.

[0220] In addition, the above recovery path prediction module (73) can store actual visit recovery results, recovery success status, recovery time, number of visits, and movement path information in the recovery history management unit (40), and continuously improve the accuracy of recovery path prediction by utilizing this as training data.

[0221] Additionally, the recovery path prediction module (73) can provide the calculated recovery possible area, recovery success probability, and path determination basis to the explainable artificial intelligence (XAI) processing unit (80), and the XAI processing unit (80) can generate the reason for selecting the recovery path and the basis for determining the priority as explanation information and provide it to the administrator terminal.

[0222] Accordingly, the above recovery path prediction module (73) can predict areas with a high probability of recovery and optimal recovery paths by comprehensively analyzing the movement patterns of rental products and the behavioral patterns of debtors beyond simple location tracking functions, and has the effect of reducing visit recovery costs and improving the recovery success rate.

[0224] The above strategy decision module (74) is configured to determine a recovery strategy for rental receivables by comprehensively analyzing the recovery risk calculated by the risk analysis module (71), the possibility of disappearance and the possibility of recovery failure calculated by the abnormal behavior risk prediction module (72), the recovery possible area and the probability of recovery success calculated by the recovery path prediction module (73), and the re-rental possibility grade and re-rental profitability calculated by the re-rental value evaluation unit (100), and may include a strategy evaluation unit (741), a recovery method determination unit (742), a priority determination unit (743), a schedule allocation unit (744), and a strategy update unit (745).

[0226] The above strategy evaluation unit (741) can calculate a Strategy Decision Index by comprehensively analyzing recovery risk, possibility of disappearance, probability of recovery success, estimated recovery cost, estimated recovery period, estimated re-rental revenue and product residual value.

[0227] In addition, the above strategy evaluation unit (741) can analyze the correlation between recovery risk and re-rental value, and can apply a priority recovery strategy to rental products that have a high recovery risk but a very high re-rental value.

[0228] Additionally, the recovery method determination unit (742) may determine a recovery execution method based on a strategy determination index. At this time, the recovery execution method may include at least one of inducing payment, inducing installment repayment, non-face-to-face recovery, visiting recovery, inducing asset return, legal action, or compulsory recovery procedures.

[0229] For example, if the delinquency period is short but the response rate to contact is high, a strategy to induce installment payments can be decided; if the delinquency period is long and there is a high possibility of disappearance, a strategy to recall visits or take legal action can be decided.

[0230] In addition, the above recovery method determination unit (742) can generate a customized recovery strategy for each debtor by analyzing the debtor's past payment history, installment repayment fulfillment rate, and consultation response rate.

[0231] In addition, the priority determination unit (743) can calculate the recovery priority for multiple rental receivables based on recovery risk, possibility of disappearance, product residual value, re-rental profitability and probability of successful recovery.

[0232] For example, if the probability of recovery is equal, products with high re-rental profitability can be selected as priority recovery targets, and if the re-rental value is equal, receivables with a high probability of being lost can be selected as priority recovery targets.

[0233] In addition, the priority determination unit (743) can classify the recovery targets into an emergency recovery group, a priority recovery group, a general recovery group, and a monitoring group.

[0234] In addition, the schedule assignment unit (744) can automatically assign a visit collection schedule, a non-face-to-face collection schedule, a legal action schedule, and a collection execution schedule based on the recovery possible area and optimal recovery route calculated by the recovery route prediction module (73).

[0235] In addition, the above schedule allocation unit (744) can automatically distribute recovery tasks by considering the location of the recovery performer, workload, number of recovery cases by region, and probability of recovery success.

[0236] In addition, the strategy decision module (74) can dynamically change the recovery strategy when new status information is received or the debtor's behavior pattern changes during the recovery process.

[0237] For example, if a debtor classified as a risk of going into hiding enters into an installment repayment agreement, the visit collection strategy can be changed to an installment repayment management strategy; conversely, if there is no response for an extended period, it can be automatically switched to a legal action strategy.

[0238] In addition, the strategy update unit (745) can continuously learn and improve the strategy decision model by collecting information on actual recovery results, recovery time, recovery success, legal action results, re-rental execution results, and recovery costs.

[0239] In addition, the strategy decision module (74) can store the determined recovery strategy, priority, schedule information, and strategy change history in the recovery history management unit (40), and the information can be recorded in a blockchain-based distributed ledger to ensure the reliability and traceability of the recovery strategy decision process.

[0240] Additionally, the strategy decision module (74) can provide the risk, possibility of evaporation, probability of successful recovery, re-rental value, and basis for priority determination used in the strategy decision to the explainable artificial intelligence (XAI) processing unit (80), and the XAI processing unit (80) can generate the reason for selecting the recovery strategy in the form of explanatory information and provide it to the administrator terminal.

[0241] Accordingly, the strategy decision module (74) can automatically determine the optimal recovery strategy by comprehensively considering the recovery risk, the possibility of disappearance, the possibility of recovery, and the asset value, and has the effect of reducing recovery costs and improving the recovery success rate and re-rental profitability.

[0243] Meanwhile, the above strategy decision module (74) may further include at least one of a legal action recommendation module (746), an installment repayment negotiation module (747), and a recovery personnel allocation module (748) for the detailed determination of the recovery strategy.

[0244] The above legal action recommendation module (746) can recommend at least one legal action plan among applying for a payment order, seizure of claims, enforcement, proceeding with litigation, or requesting a legal review by analyzing the debtor's delinquency period, delinquency amount, contact response rate, risk of recovery, and possibility of disappearance.

[0245] The above installment repayment negotiation module (747) can automatically calculate the number of installment repayments, monthly payment amount, payment deadline, and renegotiation conditions by analyzing the debtor's past payment history, income estimation information, repayment fulfillment rate, and consultation response history, and can adjust the repayment conditions according to the debtor's response results.

[0246] The above recovery personnel assignment module (748) can automatically assign recovery tasks based on the recovery possible area, recovery success probability, and visit priority information calculated by the recovery path prediction module (73), taking into account the recovery performer's current location, workload, travel distance, assigned area, and past recovery success rate.

[0247] In addition, the above-mentioned recovery personnel allocation module (748) can generate an integrated visit recovery schedule when multiple recovery targets are distributed in the same area, and can prioritize the allocation of recovery personnel to rental products classified as an emergency recovery group.

[0248] Accordingly, the strategy decision module (74) can automatically determine the optimal recovery strategy by comprehensively considering the characteristics of the debt, the behavioral patterns of the debtor, the condition of the rental product, and the possibility of recovery, and has the effect of improving the efficiency of the recovery work and the success rate of recovery.

[0250] The above-mentioned explainable artificial intelligence (XAI) processing unit (80) is configured to provide explanatory information to a manager by analyzing the basis for calculating recovery risk, potential for disappearance, potential for recovery, estimated recovery period, and recovery priority calculated by the AI ​​risk assessment unit (70), and may be configured to include at least one of a feature importance analysis module (81), a risk cause analysis module, an explanatory information generation module, a risk analysis output module (82), and a recovery history verification module (83).

[0252] The above-describeable artificial intelligence (XAI) processing unit (80) can calculate the importance of feature values ​​that affect the calculation of recovery risk by analyzing status information, behavior data, abnormal behavior score, recovery history information, and re-rental value information used in the AI ​​risk assessment unit (70). At this time, the feature values ​​may include at least one of the following: delinquency period, usage reduction rate, location change frequency, communication blockage period, long-term non-use period, contact response rate, installment repayment fulfillment rate, recovery success history, and re-rental eligibility grade.

[0253] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can analyze the major causes affecting the increase in recovery risk. For example, the impact of a decrease in product usage over the past 30 days, a long-term interruption in communication connection, repeated location changes, a decrease in the contact response rate, or an increase in the period of delinquency on the increase in recovery risk can be calculated individually.

[0254] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can calculate the risk cause ranking based on the influence of each feature value. For example, it can sort long-term non-use status, communication blockage status, increased frequency of location changes, and out-of-contact status in order of highest contribution to the increase in recovery risk and provide this to the manager.

[0255] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can compare and analyze the trend of changes in status information and the change in recovery risk for each rental product, and can track the cause of the risk by linking the point of increase in risk and the point of occurrence of abnormal behavior.

[0256] Additionally, the explainable artificial intelligence (XAI) processing unit (80) can convert the risk analysis results into visualized risk analysis information. The risk analysis information may include at least one of a risk grade graph, a feature importance graph, a graph of abnormal behavior occurrence trends, a graph predicting recovery possibilities, and a recovery priority analysis screen.

[0257] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can provide the above-mentioned risk analysis information to a manager terminal, a collection agent terminal, or a recovery performer terminal, thereby allowing the cause of the risk and the possibility of successful recovery to be verified in advance before recovery is performed.

[0258] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can evaluate the accuracy of the explanation model by comparing and analyzing the actual recovery result and the prediction result after the recovery is completed, and can update the weights or explanation rules of the explanation algorithm based on the evaluation result.

[0259] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can provide recovery risk, explanation information, feature importance analysis results, abnormal behavior analysis results, recovery strategy decision information, and actual recovery result information to the recovery history verification module (83), and the information can be stored in a blockchain-based distributed ledger and used as verification data for the appropriateness of future recovery procedures and decision-making processes.

[0260] In addition, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can calculate the relative risk level within the same product family, same region, or same delinquency period group by comparing and analyzing the risk analysis results for multiple rental receivables, and can improve the objectivity of determining the recovery priority based on this.

[0261] Accordingly, the above-mentioned explainable artificial intelligence (XAI) processing unit (80) can intuitively provide the basis for calculation of the judgment result of the AI ​​risk assessment unit (70) to the manager, and has the effect of improving the reliability of establishing recovery strategies and ensuring transparency and explainability of AI-based decision-making.

[0263] Meanwhile, the above-mentioned feature importance analysis module (81) is configured to generate a basis for calculating risk by analyzing the influence of feature values ​​used in the process of calculating recovery risk, potential for disappearance, potential for recovery failure, and recovery priority calculated by the AI ​​risk assessment unit (70), and may include a feature value extraction unit (811), an influence calculation unit (812), a contribution analysis unit (813), and an explanation data generation unit (814).

[0265] The above feature value extraction unit (811) can collect status information including power supply status, usage frequency, usage time, location information, location movement frequency, movement distance, communication connection status, communication blockage period, maintenance history, and product abnormality status collected from the IoT status collection unit (50), and behavioral data including delinquency period, delinquency amount, payment history, contact response rate, installment repayment fulfillment rate, and application access history provided from the debt management unit (20).

[0266] Additionally, the feature value extraction unit (811) can extract abnormal behavior data including abnormal behavior scores, long-term non-use status, location deviation status, rapid decrease in usage status, power cut-off status, and communication cut-off status provided by the abnormal behavior analysis unit (60).

[0267] In addition, the impact calculation unit (812) can quantitatively calculate the impact of each feature value on the recovery risk for a plurality of feature values ​​input into the artificial intelligence model of the AI ​​risk assessment unit (70). At this time, the impact calculation unit may use at least one explanation algorithm among SHAP (Shapley Additive Explanations), LIME (Local Interpretable Model-Agnostic Explanations), feature contribution analysis technique, rule-based analysis technique, or decision tree-based analysis technique.

[0268] For example, it is possible to calculate the impact on recovery risk as the long-term non-use period increases, and the impact contributing to the reduction of recovery risk as the contact response rate increases.

[0269] In addition, the above-mentioned impact calculation unit (812) can analyze the correlation between multiple feature values ​​as well as individual feature values. For example, when a long-term non-use state and a communication blocking state occur simultaneously, a higher risk contribution than that of a single feature value can be calculated.

[0270] In addition, the contribution analysis unit (813) can calculate the extent to which each feature value contributed to the recovery risk, potential for evasion, potential for recovery failure, and recovery priority in the form of a percentage or score.

[0271] For example, among the causes of increased recovery risk, it can be analyzed that long-term non-use status contributes 35%, communication blockage status 25%, increased frequency of location change 20%, increased period of delinquency 15%, and decreased response rate to contact 5%.

[0272] In addition, the contribution analysis unit (813) can analyze risk-increasing factors and risk-reducing factors separately, and can classify feature values ​​that contribute to improving the probability of successful recovery and feature values ​​that contribute to increasing the probability of failure of recovery, respectively.

[0273] In addition, the above explanation data generation unit (814) can generate explanation information based on the results of impact analysis and contribution analysis. For example, it can generate information on the causes of risk such as "80% decrease in usage over the last 30 days," "no communication connection over the last 15 days," "moved 120 km relative to the installation location," and "no response to contact over the last 60 days."

[0274] In addition, the above explanation data generation unit (814) can generate an importance ranking for each feature value and can provide the top feature value that has the greatest influence on the increase in recovery risk to the manager.

[0275] In addition, the feature importance analysis module (81) can compare and analyze the importance of feature values ​​for the same product family, same region, or same delinquency period group, and can derive the difference in features between successful recovery cases and failed recovery cases.

[0276] In addition, the feature importance analysis module (81) can provide the calculated feature importance information and explanatory data to the risk analysis output module (82) to visualize them in the form of a graph, chart, priority list, or dashboard.

[0277] In addition, the feature importance analysis module (81) may provide the result of the feature value importance analysis, the basis for risk calculation, and explanatory information to the recovery history verification module (83), and the information may be stored in a blockchain-based distributed ledger and used as verification data to ensure transparency and traceability of the AI ​​decision-making process.

[0278] Accordingly, the above-mentioned feature importance analysis module (81) can quantitatively analyze the basis for the recovery risk calculation process of the AI ​​risk assessment unit (70), and by supporting the manager to intuitively identify the cause of the risk occurrence, it has the effect of improving the reliability and explainability of the recovery strategy decision.

[0280] Meanwhile, the above-mentioned risk analysis output module (82) is configured to provide the explanatory information generated by the feature importance analysis module (81) and the explanatory data generation unit (814) to a manager terminal, a collection agent terminal, or a recovery operator terminal by visualizing it, and may be configured to include a risk grade display unit (821), a risk factor visualization unit (822), a status information display unit (823), a recovery strategy display unit (824), and a history analysis display unit (825).

[0281] The above risk grade display unit (821) can display a risk grade for each rental receivable based on the recovery risk, possibility of disappearance, possibility of failure to recover, and recovery priority calculated by the AI ​​risk assessment unit (70). At this time, the risk grade may be displayed as a low-risk group, caution group, high-risk group, and emergency recovery group, but is not limited thereto.

[0282] In addition, the risk factor visualization unit (822) can display the contribution of the feature value calculated by the feature importance analysis module (81) in the form of a graph, chart, heatmap, or ranking list. For example, it can be visualized that a long-term non-use state has an impact of 40%, a communication blockage state has an impact of 30%, an increase in location movement frequency has an impact of 20%, and an increase in delinquency period has an impact of 10%.

[0283] In addition, the above risk factor visualization unit (822) can distinguish between risk-increasing factors and risk-reducing factors and display the main causes that have the greatest impact on the recovery risk first.

[0284] In addition, the status information display unit (823) can display the current location, usage status, communication connection status, power status, usage frequency, maintenance status, and whether abnormal behavior has occurred of the rental product provided by the IoT status collection unit (50) in real time or periodically.

[0285] In addition, the status information display unit (823) can display location information of the rental product on a map-based screen and can also display recent location movement paths, major areas of stay, and areas where recovery is possible.

[0286] Additionally, the recovery strategy display unit (824) may display a recovery strategy generated by the strategy decision module (74) and the autonomous recovery strategy generation unit (90). In this case, the recovery strategy may include at least one of the following: inducing installment repayment, non-face-to-face recovery, visit recovery, legal action, forced recovery, or re-rental priority recovery strategy.

[0287] In addition, the recovery strategy display unit (824) can display the recovery priority, expected recovery date, recovery success probability, expected recovery cost, and expected re-rental profit together.

[0288] In addition, the history analysis display unit (825) can display the trend of change in recovery risk, the trend of change in abnormal behavior score, the status of delinquency, and the history of changes in recovery strategy in the form of a time series graph.

[0289] In addition, the history analysis display unit (825) can display the results of a comparative analysis of past recovery success cases and recovery failure cases with the current status of rental receivables, and can also provide the recovery success rate and recommended recovery strategy for similar cases.

[0290] In addition, the above-mentioned risk analysis output module (82) can output risk alert information in the form of a warning message, notification window, text message, or mobile notification when a risk condition exceeding a standard value set by the manager occurs.

[0291] In addition, the risk analysis output module (82) can provide the risk analysis results, recovery strategy, explanatory information, and manager action history to the recovery history management unit (40) and the recovery history verification module (83), and the information can be stored in a blockchain-based distributed ledger to ensure traceability and reliability of the AI ​​decision-making results.

[0292] Accordingly, the above-mentioned risk analysis output module (82) can provide AI risk assessment results and explanatory information in a form that an administrator can intuitively understand, and has the effect of improving the speed and accuracy of establishing recovery strategies and increasing the efficiency of recovery operations.

[0294] The above recovery history verification module (83) is configured to verify whether recovery history information, risk analysis information, explanation information, and recovery strategy information generated during the recovery process of rental receivables are falsified or altered, and to ensure the reliability of the recovery procedure, and may be configured to include a history collection unit (831), a hash generation unit (832), a blockchain record unit (833), a verification processing unit (834), and a history tracking unit (835).

[0295] The above history collection unit (831) can collect recovery-related information generated from the debt information collection unit (10), debt recovery processing unit (30), recovery history management unit (40), AI risk assessment unit (70), explainable artificial intelligence (XAI) processing unit (80), autonomous recovery strategy generation unit (90), and non-face-to-face recovery support unit (110).

[0296] The above-mentioned recovery-related information may include, but is not limited to, recovery risk, possibility of disappearance, probability of recovery success, recovery priority, recovery strategy, explanatory information, information on on-site recovery execution, information on non-face-to-face collection execution, information on legal action execution, product return information, recovery completion information, and re-rental execution information.

[0297] Additionally, the hash generation unit (832) can generate a hash value for the collected recovery-related information. At this time, the hash algorithm may use at least one of SHA-256, SHA-3, Keccak, Blake2, or similar cryptographic hash algorithms.

[0298] In addition, the hash generation unit (832) can generate a hash value by linking a unique identifier for each recovery history information with the creation time information, and thereby check whether the same information is registered multiple times.

[0299] Additionally, the blockchain record (833) can record the generated hash value and recovery history information in a blockchain-based distributed ledger. In this case, the distributed ledger can be stored in a distributed manner across multiple management servers or participating nodes, and each node can share the same recovery history data.

[0300] In addition, the blockchain record (833) can record the AI ​​risk assessment results, recovery strategy decision results, and explanation information generation results in chronological order on the blockchain, thereby ensuring traceability of the recovery strategy decision process.

[0301] In addition, the verification processing unit (834) can verify whether the data has been changed by comparing the hash value of the current data with the hash value stored in the blockchain when the recovery history information is modified or retrieved.

[0302] For example, if the recovery completion date, information regarding the person in charge of recovery, or the results of legal action are arbitrarily changed, the fact of tampering can be detected through a discrepancy with the stored hash value.

[0303] In addition, the verification processing unit (834) can verify the relationship between the risk level calculated by the AI ​​risk assessment unit (70) and the actual recovery performance results, and can track which risk analysis results a specific recovery strategy was determined on.

[0304] In addition, the verification processing unit (834) can verify the correlation between the explanatory information generated by the explainable artificial intelligence (XAI) processing unit (80) and the actual risk calculation result, and can ensure transparency in the AI ​​decision-making process.

[0305] Additionally, the history tracking unit (835) may be configured to allow viewing the history of recovery progress for a specific rental product or a specific receivable in chronological order. At this time, the time of receivable occurrence, the time of delinquency occurrence, the time of risk increase, the time of recovery strategy change, the time of legal action execution, and the time of recovery completion can be tracked in conjunction.

[0306] In addition, the history tracking unit (835) can record the history viewing and modification requests made by the recovery manager, collection manager, legal manager, or manager, and thereby perform an audit function of the internal management process.

[0307] In addition, the above recovery history verification module (83) can establish a single verification system for the entire process of rental receivable management by linking and storing the recovery strategy change history, non-face-to-face recovery support history, installment repayment negotiation history, visit recovery performance history, and re-rental processing history in a blockchain-based distributed ledger.

[0308] In addition, the above recovery history verification module (83) can generate verification information by comparing and analyzing the actual recovery results and the AI ​​prediction results, and the verification information can be used as training data for the AI ​​risk assessment unit (70) to improve prediction accuracy.

[0309] Accordingly, the above recovery history verification module (83) can record and verify the entire process from recovery risk analysis to recovery completion and re-rental execution on a blockchain-based distributed ledger, and has the effect of preventing falsification of the recovery history and ensuring transparency and reliability of the AI-based recovery decision-making process.

[0311] The above-mentioned autonomous recovery strategy generation unit (90) is configured to automatically generate an optimal recovery strategy by comprehensively analyzing the recovery risk, possibility of disappearance, possibility of recovery failure, probability of recovery success, and recovery priority calculated by the AI ​​risk assessment unit (70), and the re-rental possibility grade, re-rental profitability, and product residual value calculated by the re-rental value assessment unit (100). It may be configured to include a strategy analysis module (91), a recovery plan generation module (92), a priority determination module (93), a schedule management module (94), and a strategy learning module (95).

[0313] The above strategy analysis module (91) can analyze recovery risk, possibility of disappearance, possibility of recovery failure, possibility of product concealment, and information on recoveryable areas provided by the AI ​​risk assessment unit (70).

[0314] In addition, the above strategy analysis module (91) can generate a strategy decision index necessary for determining a recovery strategy by analyzing the debtor's payment history, contact response rate, installment repayment fulfillment rate, application access history, location change history, and abnormal behavior occurrence history.

[0315] The above recovery plan generation module (92) can generate at least one recovery strategy among a payment inducement strategy, an installment repayment inducement strategy, a non-face-to-face recovery strategy, a visit recovery strategy, a product return inducement strategy, a legal action strategy, or a forced recovery strategy based on the above strategy decision index.

[0316] For example, if the recovery risk is low and the contact response rate is high, a strategy to induce installment repayment can be created; if the recovery risk is high and the likelihood of disappearance is high, a strategy for on-site recovery or legal action can be prioritized.

[0317] In addition, the above recovery plan generation module (92) can generate a recovery strategy by considering both the recovery risk and the profitability of re-rental. For example, for rental products that have a high recovery risk but very high re-rental value, an emergency recovery strategy can be generated.

[0318] The above priority determination module (93) can calculate the recovery priority for multiple rental receivables based on recovery risk, product residual value, re-rental profitability, delinquency period and recovery success probability.

[0319] In addition, the priority determination module (93) can classify claims into an emergency recovery group, a priority recovery group, a general recovery group, and an observation group, and can apply different recovery strategies to each group.

[0320] The above schedule management module (94) can automatically generate a visit recovery schedule, a collection schedule, a legal action schedule, and a non-face-to-face recovery schedule based on the recovery possible area and optimal recovery route calculated by the recovery route prediction module (73).

[0321] In addition, the above schedule management module (94) can automatically assign recovery tasks by considering the recovery performer's current location, assigned area, workload, and probability of recovery success.

[0322] The above strategy learning module (95) can continuously learn a recovery strategy generation model by collecting information on actual recovery results, whether recovery was successful, recovery time, legal action results, installment repayment results, re-rental results, and recovery costs.

[0323] In addition, the strategy learning module (95) can improve the accuracy of strategy generation for similar bonds in the future by analyzing the success rate and failure rate of a specific strategy.

[0324] Additionally, the above-mentioned autonomous recovery strategy generation unit (90) can provide the generated recovery strategy to the non-face-to-face recovery support unit (110) and the debt recovery processing unit (30), and the above-mentioned non-face-to-face recovery support unit (110) can automatically transmit information to the debtor regarding payment guidance, installment repayment proposals, recovery requests, or planned legal action.

[0325] In addition, the autonomous recovery strategy generation unit (90) can provide the basis for generating the recovery strategy, the history of strategy changes, and the results of strategy execution to the recovery history management unit (40) and the recovery history verification module (83), and the information can be stored in a blockchain-based distributed ledger to ensure the reliability and traceability of the strategy decision process.

[0326] Accordingly, the above-mentioned autonomous recovery strategy generation unit (90) can automatically generate an optimal recovery strategy by comprehensively analyzing the recovery risk, debtor behavior pattern, product status information and re-rental value, and has the effect of improving the recovery success rate, reducing recovery costs, and maximizing the profitability of rental assets.

[0328] The above re-rental value evaluation unit (100) is configured to calculate whether re-rental is possible, re-rental possible grade, expected re-rental profit, and asset utilization value by analyzing the condition information, maintenance history, usage history, market demand information, and recovery cost information of the recovered rental product, and may be configured to include a product condition analysis module (101), a residual value calculation module (102), a re-rental profitability analysis module (103), and a recycling decision module (104).

[0330] The above product condition analysis module (101) can collect product usage years, cumulative usage time, usage frequency, maintenance performance history, filter replacement history, repair history, parts replacement history, and product malfunction history provided by the IoT condition collection unit (50), the receivables management unit (20), and the recovery history management unit (40).

[0331] In addition, the product condition analysis module (101) can calculate a product condition grade by analyzing the external condition, contamination condition, part wear condition, performance degradation condition, and damage of the recovered rental product. At this time, the product condition grade may be classified into Grade A, Grade B, Grade C, or unusable grade, but is not limited thereto.

[0332] In addition, the product status analysis module (101) can calculate the actual usage intensity by analyzing the operation history and IoT status information stored in the rental product. For example, in the case of a water purifier, it can analyze the cumulative water intake, filter usage, and operating time, and in the case of an air purifier, it can analyze the cumulative operating time, fan operating time, and remaining filter life.

[0333] The above residual value calculation module (102) can calculate the residual value of a product by analyzing the manufacturing date, installation date, recovery date, years of use, depreciation information, and intensity of use of the rental product. At this time, the residual value can be calculated by reflecting the depreciation rate by product type, maintenance status, degree of performance degradation, and market transaction price.

[0334] The above re-rental profitability analysis module (103) can calculate the product condition, parts replacement cost, cleaning cost, restoration cost, logistics cost, inspection cost and maintenance cost of the recovered rental product.

[0335] In addition, the above-mentioned re-rental profitability analysis module (103) can calculate a market demand index by analyzing regional rental demand, contract signing rate by product group, market inventory level, seasonal demand change, customer preference, past re-rental success history and used transaction history by model.

[0336] In addition, the above-mentioned re-rental profitability analysis module (103) can calculate the expected re-rental profit by comprehensively analyzing the expected re-rental period, the expected monthly rental fee, the expected contract maintenance period, the expected maintenance costs, and the expected recovery costs.

[0337] In addition, the above-mentioned re-rental profitability analysis module (103) can calculate re-rental profitability by comprehensively analyzing residual value, market demand index, restoration cost and expected re-rental profit, and can generate a re-rental profitability index (Rental Profitability Index).

[0338] In addition, the above-mentioned re-rental profitability analysis module (103) can calculate the probability of future re-rental success using an artificial intelligence-based prediction model, and the prediction model can be built by learning past re-rental success cases, contract maintenance period, product condition grade, customer preference, and market demand data.

[0339] In addition, the above-mentioned re-rental profitability analysis module (103) can adjust the recovery priority by comparing the expected re-rental profit relative to the recovery cost. For example, among multiple rental products with similar recovery potential, a priority recovery strategy can be assigned to the product with the higher expected re-rental profit.

[0340] The recycling decision module (104) can determine a method for utilizing the recovered rental product by comparing and analyzing the re-rental profitability index calculated by the re-rental profitability analysis module (103), the recovery cost, the restoration cost, the residual value, and the product condition grade.

[0341] In addition, the recycling decision module (104) can classify a re-rental target if the re-rental profitability is above a threshold value, and can classify a used item for sale if the re-rental profitability is low but the used market value is above a threshold value.

[0342] In addition, the recycling decision module (104) can classify a product as a product for recycling if it is difficult to reuse the entire product but the value of the core component is high, and can decide to discard a product if both the residual value and the reuse value of the product are below a standard value.

[0343] In addition, the recycling decision module (104) can provide the decision results regarding whether to re-rent, whether to sell as used, whether to recycle or dispose of parts to the autonomous recovery strategy generation unit (90) and the strategy decision module (74), and can be used to determine the recovery priority and establish a recovery strategy.

[0344] Additionally, the re-rental value evaluation unit (100) can provide the re-rental processing result, whether the re-rental contract is concluded, information on the generation of re-rental revenue, the result of used sales, the result of parts recycling, and the asset utilization history to the recovery history management unit (40), and the information can be used as training data for the AI ​​risk evaluation unit (70) and the strategy decision module (74).

[0345] In addition, the above-mentioned re-rental value evaluation unit (100) can calculate the economic value of the entire product lifecycle by analyzing the trend of changes in the asset value of the rental product, and thereby support the establishment of asset management strategies for the rental business operator.

[0346] Accordingly, the above-mentioned re-rental value evaluation unit (100) can quantitatively evaluate the reuse value, re-rental profitability, and asset utilization potential of the recovered rental product beyond simply determining whether to recover the receivable, and by determining the optimal utilization plan for the recovered rental product, it has the effect of improving the asset utilization and profitability of the rental business operator.

[0348] The above-mentioned non-face-to-face recovery support unit (110) is configured to improve the recovery rate of rental receivables by interacting with the debtor in a non-face-to-face manner according to the recovery strategy generated by the AI ​​risk assessment unit (70) and the autonomous recovery strategy generation unit (90), and may be configured to include at least one of an information generation module, a multi-channel transmission module, a product linkage guidance module, an electronic agreement processing module, and a recovery support management module.

[0350] The above non-face-to-face recovery support unit (110) can generate payment guidance information, delinquency warning information, installment repayment proposal information, recovery schedule information, visit recovery guidance information, legal action schedule information, and product recovery request information based on debtor information, delinquency information, and recovery progress information provided by the debt management unit (20).

[0351] In addition, the above non-face-to-face recovery support unit (110) can transmit the information to the debtor's terminal using at least one communication means among text message (SMS), multimedia message (MMS), email (E-mail), mobile application (App), KakaoTalk notification, push notification, automatic voice guidance (ARS), or chatbot service.

[0352] In addition, the above-mentioned non-face-to-face recovery support unit (110) can determine the optimal transmission time and transmission method based on the recovery risk, response probability, and contact success probability calculated by the AI ​​risk assessment unit (70). For example, recovery guidance information can be transmitted preferentially during a specific time period to debtors with a high response rate during that time period.

[0353] In addition, the above-mentioned non-face-to-face collection support unit (110) can output collection-related information using a display unit, indicator light, speaker, voice output device, or mobile linkage interface provided in the rental product. For example, overdue notice information or collection schedule information can be displayed on the display screen of a water purifier, or collection-related guidance voice can be output through the voice output device of a massage chair.

[0354] In addition, the above-mentioned non-face-to-face recovery support unit (110) may restrict some functions of the rental product or restrict the use of the service in stages when the delinquency period or the recovery risk level is above a threshold value. For example, it may restrict additional functions of a water purifier or restrict specific operating modes of an air purifier, and this can be used as a means to make the debtor aware of the fact that the recovery procedure is in progress.

[0355] In addition, the above-mentioned non-face-to-face recovery support unit (110) may provide an electronic agreement interface so that the debtor can apply for installment repayment, apply for a change in payment schedule, negotiate a recovery schedule, apply for product return, or request consultation through a terminal. At this time, the expression of intent entered by the debtor can be verified through an electronic signature or identity verification procedure.

[0356] In addition, the above non-face-to-face recovery support unit (110) can collect and store behavioral data regarding whether the debtor responds, the response time, whether the message is viewed, whether the link is accessed, whether an electronic agreement is concluded, and whether payment is made. The above data can be used as learning data for the AI ​​risk assessment unit (70) and as analysis data for improving recovery strategies.

[0357] In addition, the above-mentioned non-face-to-face recovery support unit (110) can be linked with a chatbot or generative AI-based counseling module to automatically provide the debtor with information regarding the status of delinquency, payment method, installment repayment conditions, product return procedure, and recovery schedule. Accordingly, continuous communication with the debtor is possible without the intervention of counseling personnel.

[0358] In addition, the above-mentioned non-face-to-face recovery support unit (110) can change the recovery strategy by analyzing the debtor's response results in real time. For example, if the debtor does not respond to the installment repayment proposal, the priority for visit recovery can be raised or legal action procedures can be initiated.

[0359] In addition, the above-mentioned non-face-to-face recovery support unit (110) can transmit transmitted guidance information, debtor's response information, electronic agreement conclusion information, recovery schedule consultation information, and product return request information to the recovery history management unit (40), and the above-mentioned information can be stored in a blockchain-based distributed ledger and used as data to verify the legality and execution history of the recovery procedure in the future.

[0360] In addition, the above-mentioned non-face-to-face collection support unit (110) can automatically generate and provide product return location guidance, unmanned collection box guidance, or courier collection guidance information based on the location information, usage status, and scheduled collection schedule of the rental product. Accordingly, various non-face-to-face collection routes can be supported in addition to visit collection.

[0361] Accordingly, the above-mentioned non-face-to-face recovery support unit (110) can automate the recovery process through non-face-to-face interaction with the debtor, reduce the operating costs of the collection personnel, improve the recovery success rate, and have the effect of digitizing the entire process of rental receivable recovery.

[0363] As described above, the present invention has been explained by specific details such as specific components, limited embodiments, and drawings; however, this is provided merely to aid in a more comprehensive understanding of the invention, and the invention is not limited to the above embodiments. A person skilled in the art can make various modifications and variations from this description.

[0365] Accordingly, the scope of the present invention should not be limited to the described embodiments, and all things equivalent to or having equivalent variations to the claims set forth below, as well as the claims set forth below, shall be considered to fall within the scope of the concept of the present invention. Explanation of the symbols

[0367] 1 : Rental Receivables Management System 10 : Bond Information Collection Department 20 : Debt Management Department 30 : Debt Collection Processing Unit 40: Recovery History Management Department 50: IoT State Collection Unit 60 : Abnormal Behavior Analysis Department 70 : AI Risk Assessment Department 71: Risk Analysis Module 72: Abnormal Behavior Risk Prediction Module 73: Recovery Path Prediction Module 74: Strategy Decision Module 80: Explainable AI Processor 81: Feature Importance Analysis Module 82: Risk Analysis Output Module 83 : Recovery History Verification Module 90 : Autonomous Recovery Strategy Generation Unit 91: Strategy Analysis Module 92: Recovery Plan Generation Module 93: Priority Determination Module 94: Schedule Management Module 95 : Strategy Learning Module 100 : Re-rental Value Assessment Department 101: Product Condition Analysis Module 102: Residual Value Calculation Module 103: Re-rental Profitability Analysis Module 104: Recycling Decision-making Module 110 : Non-face-to-face Collection Support Department

Claims

Claim 1 A receivable information collection unit (10) that collects information on rental receivables generated due to overdue rental fees of a rental product; a receivable management unit (20) that stores and manages debtor information, rental product information, and overdue information corresponding to the rental receivables; a receivable recovery processing unit (30) that performs legal action, collection, installment repayment, or receivable recovery procedures depending on the recoverability of the rental receivables; and a recovery history management unit (40) that stores the recovery results and recovery history of the rental receivables; and an IoT status collection unit (50) that collects at least one status information from the rental product among the product's communication connection status, power application status, operating time, usage frequency, sensor operation information, installation location information, network connection status, maintenance history, filter replacement history, and product abnormal status; and an abnormal behavior that generates a standard model for the normal usage pattern of the rental product and determines whether there is abnormal behavior by analyzing the usage pattern change rate, long-term non-use status, power cut-off duration, location deviation, communication cut-off status, and rapid decrease in usage by comparing with the status information. Analysis unit (60); an AI risk assessment unit (70) that calculates whether the rental product is actually used, the possibility of the debtor going into hiding, the possibility of recovering the rental product, the estimated recovery period, and the recovery priority based on at least one behavioral data among the above status information, abnormal behavior analysis results, rental fee delinquency information, the debtor's payment history, contact response history, application access history, and location change history; an explainable artificial intelligence (XAI) processing unit (80) that analyzes the influence of feature values ​​that serve as the basis for calculating the recovery risk calculated by the AI ​​risk assessment unit (70), and generates explanatory information on the impact of long-term non-use status, frequency of location change, communication blocking status, usage reduction rate, and delinquency progress status on the recovery risk; and based on the possibility of recovery, the estimated recovery period, and the residual value of the rental product calculated by the AI ​​risk assessment unit (70), induces visit recovery, legal action, and installment repayment,The system further comprises: an autonomous recovery strategy generation unit (90) that automatically generates a recovery strategy including collection priority and collection schedule assignment; a re-rental value evaluation unit (100) that calculates whether the rental product is reusable, a re-rental eligibility grade, and expected re-rental revenue by analyzing the rental product's usage years, usage frequency, maintenance history, presence of exterior damage, condition of parts wear, used transaction history by model, and re-rental history; and a non-face-to-face recovery support unit (110) that outputs payment guidance information, recovery schedule information, installment repayment proposal information, product recovery request information, or service restriction guidance information to the debtor terminal or rental product according to the recovery strategy generated by the autonomous recovery strategy generation unit (90); wherein the AI ​​risk evaluation unit (70) includes a risk analysis module (71) that calculates the recovery risk by analyzing the rental product's usage pattern change rate, location movement frequency, duration of long-term non-use, communication blocking frequency, and delinquency progress status as time-series data based on the status information and behavioral data, and a standard model of the rental product's normal usage pattern and current The system includes an abnormal behavior risk prediction module (72) that calculates an abnormal behavior score by calculating the deviation between state information and predicts the possibility of disappearance and the possibility of recovery failure based on the abnormal behavior score, a recovery path prediction module (73) that calculates the recovery possible area, visit recovery priority, and expected recovery path of the rental product based on the location information, usage status, and past recovery success history of the rental product, and a strategy decision module (74) that calculates the recovery strategy priority by synthesizing the recovery risk, abnormal behavior score, recovery possible area, and re-rental possibility grade, and the explainable artificial intelligence (XAI) processing unit (80) includes a feature importance analysis module (81) that calculates the influence of the feature value used in the risk analysis module (71) and the abnormal behavior risk prediction module (72), and generates explanatory information on the contribution of long-term non-use status, frequency of location change, communication blocking status, usage reduction rate, and delinquency progress status to the increase in recovery risk.A rental receivables management system comprising: a risk analysis output module (82) that converts the actual usage status, long-term non-use status, out-of-location status, recovery risk status, and recovery priority of a rental product into visualized risk analysis information and outputs it to an administrator terminal based on the above-mentioned explanatory information; a recovery history verification module (83) that verifies whether the rental product recovery procedure has been tampered with by storing the above-mentioned explanatory information, recovery risk level, recovery power, rental product location change information, visit recovery execution information, recovery completion information, product handover information, and legal action execution information in a blockchain-based distributed ledger; and a re-rental value evaluation unit (100) comprising a re-rental profitability analysis module (103) that calculates re-rental profitability by analyzing the product condition of the recovered rental product, parts replacement costs, cleaning or restoration costs, expected re-rental period, and market demand by model, and a recycling decision module (104) that determines whether to prioritize recovery, whether to re-rent, whether to sell as used, or whether to dispose of by comparing the re-rental profitability and recovery costs. Claim 2 delete Claim 3 delete

Images