Search

CN-122023033-A - Financial customer data management method and system based on big data evaluation

CN122023033ACN 122023033 ACN122023033 ACN 122023033ACN-122023033-A

Abstract

The invention relates to the technical field of financial client data management, in particular to a financial client data management method and system based on big data evaluation, which is used for analyzing cargo pay triggering potential energy of a financial client by combining principal attribute data and historical financial transaction data of the financial client; the method comprises the steps of combining cargo transportation data and transportation route data of related cargoes of financial clients to analyze abnormal cargo state risks of the current transportation tasks of the financial clients, evaluating the risk of paying the current transportation tasks of the financial clients based on cargo paying triggering potential energy analysis results of the financial clients and cargo state abnormal risk analysis results of the current transportation tasks of the financial clients, and carrying out early warning of cargo paying based on the risk evaluation results of paying the current transportation tasks of the financial clients, so that accuracy of risk pricing and settlement and checking is improved, potential cargo loss and paying cost are effectively reduced, and customer service experience and risk management and control efficiency are improved.

Inventors

  • HUANG JIANGYI
  • YAO JING
  • YANG YANG

Assignees

  • 苏州大学

Dates

Publication Date
20260512
Application Date
20260413

Claims (7)

  1. 1. The financial customer data management method based on big data evaluation is characterized by comprising the following steps: s1, acquiring main body attribute data and historical financial transaction data of a financial customer serving as an applicant, and simultaneously acquiring cargo transportation data and transportation route data of related cargoes of the financial customer serving as the applicant; S2, extracting a task number, a cargo type code, a damage position code, a damage area occupation ratio, an average depth change value of a damage part and a binary mark of whether a corresponding transportation task generates a claim according to a completed historical transportation task record in historical financial transaction data, extracting historical claim settlement data sets of corresponding clients in main attribute data, constructing a historical damage claim settlement data set, calculating the percentage of the number of records of claim settlement under each cargo type code and damage position code in the total number of records in a statistics mode based on the historical damage claim settlement data sets, and defining the percentage as a reference function influence weight corresponding to each cargo type code and damage position code in a pairwise mode; S3, dividing the total duration of each transportation task in the historical cargo state normal data set into a plurality of time windows, extracting position change parameters, speed fluctuation parameters, driving excitation parameters, environment deviation parameters, sealing state parameters and mechanical attenuation state risk coefficients representing the self mechanical property attenuation of the transportation carrier from cargo transportation data for each time window as transportation state parameters representing each time window, training a cargo state abnormal risk detection model based on a nonlinear coupling relation of the transportation state parameters of a multidimensional time sequence by utilizing the transportation state parameters of each time window in the historical cargo state normal data set, and judging the probability of occurrence of unauthorized path deviation abnormal risk in the corresponding time window according to comprehensive analysis of five transportation state parameters, and analyzing the cargo state abnormal risk of the current transportation task of a financial client based on the probability of occurrence of unauthorized path deviation abnormal risk in the corresponding time window, and the historical reference risk coefficient matched with each time window in the transportation route data output by the cargo state abnormal risk detection model; The position change parameter is used for identifying position change abnormality to find path deviation risk, the speed fluctuation parameter and the driving excitation parameter are used for identifying driving behavior, the environment deviation parameter is used for judging whether cargo storage conditions are in compliance, the sealing state parameter is used for detecting whether cargo is illegally contacted, and the mechanical attenuation state risk coefficient representing the self mechanical property attenuation of the transport carrier is used for identifying potential transport interruption or cargo loss risk caused by technical condition degradation of the transport carrier; S4, inputting the cargo pay triggering potential energy analysis result of the financial client and the cargo state abnormal risk analysis result of the current transportation task of the financial client into a preset risk fusion evaluation function based on a dynamic game theory, and evaluating the pay risk of the current transportation task of the financial client; S5, carrying out early warning on the goods of the financial client serving as the insuring party based on the risk assessment result of the payment of the current transportation task of the financial client.
  2. 2. The method for managing financial customer data based on big data evaluation according to claim 1, wherein the step S2 is to construct a cargo pay trigger potential prediction model by using the historical damage claim data set and the reference function influence weight, and analyze the cargo pay trigger potential of the financial customer, and the method comprises the following steps: s23, constructing and training a cargo pay triggering potential energy prediction model based on a logistic regression model and the historical damage claim settlement data set; S24, aiming at the current transported goods of the financial customer, acquiring a damaged position code, a damaged area occupation ratio and an average depth change value of a damaged part of the current transported goods through image recognition and three-dimensional scanning, acquiring a goods type code of the current transported goods, inquiring to obtain reference function influence weights corresponding to the corresponding goods type code and the damaged position code in pairs according to the goods type code and the damaged position code, acquiring historical claims rate of the corresponding customer, inputting the damaged area occupation ratio, the average depth change value of the damaged part, the corresponding reference function influence weights, the historical claims rate of the corresponding customer and the goods type code of the current transported goods into a trained goods claims triggering potential energy prediction model, outputting the triggering claims probability value of the current transported goods by the model, and taking the triggering claims probability value of the current transported goods as the goods claims triggering potential energy of the current transported goods of the financial customer.
  3. 3. The method according to claim 2, wherein the step S3 of extracting the position change parameter, the speed fluctuation parameter, the driving shock parameter, the environment deviation parameter, the sealing state parameter, and the mechanical attenuation state risk coefficient representing the attenuation of the mechanical performance of the transportation vehicle as the transportation state parameter representing each time window from the cargo transportation data comprises the following steps: S31, acquiring a vehicle longitude and latitude coordinate sequence and an instantaneous speed sequence of a transportation vehicle corresponding to a transportation task from cargo transportation data, a longitudinal acceleration event signal sequence and a transverse acceleration event signal sequence which are acquired through a vehicle controller local area network bus and exceed a preset intensity threshold, a temperature reading sequence and a relative humidity reading sequence which are acquired through an environment sensor fixed in a cargo compartment, an electronic seal state signal sequence which are acquired through an electronic seal installed at a freight unit door, and an original diagnosis data stream which is acquired in real time through a vehicle self-diagnosis system interface and reflects the working state of each subsystem of the vehicle, wherein the electronic seal state signal comprises a locking state signal and an unlocking state signal; S32, dividing the time into a plurality of time windows with equal time intervals according to the duration of the transportation task, and acquiring cargo transportation data and transportation route data corresponding to each time window in the transportation task; s33, calculating a spherical linear displacement distance according to longitude and latitude coordinates of the carrier at the beginning and the end of the corresponding time window, and taking the spherical linear displacement distance as a position change parameter of the corresponding time window; S34, calculating standard deviations of all instantaneous speed reading sequences in the corresponding time window, and taking the standard deviations as speed fluctuation parameters of the corresponding time window; S35, accumulating and counting the total times of the longitudinal acceleration event signal and the transverse acceleration event signal triggered in the corresponding time window, and taking the total times as driving excitation parameters of the corresponding time window; S36, judging whether the reading in any reading sequence of the temperature reading sequence and the relative humidity reading sequence in the corresponding time window exceeds a preset safety threshold range, if yes, assigning the environment deviation parameter of the corresponding time window as 1, otherwise, assigning the environment deviation parameter of the corresponding time window as 0; s37, identifying whether a jump from a locking state signal to an unlocking state signal occurs in the electronic seal state signal sequence in the corresponding time window, if so, assigning a sealing state parameter of the corresponding time window to be 1, otherwise, assigning 0; S38, extracting features of the original diagnostic data streams which are collected in the corresponding time windows and reflect the working states of all subsystems of the vehicle, and calculating the mechanical property attenuation degree of the transport vehicle in the current time window based on the extracted feature values to be used as a mechanical attenuation state risk coefficient for representing the mechanical property attenuation of the transport vehicle.
  4. 4. A method for managing financial customer data based on big data evaluation according to claim 3, wherein the step S3 of analyzing abnormal risk of cargo state of the current transportation task of the financial customer comprises the following steps: s39, constructing a transport state parameter coupling analysis model based on a deep neural network, taking transport state parameter vectors of each time window in a historical cargo state normal data set as input, and training with a reconstructed corresponding vector as a target, so that the model can learn nonlinear coupling relations among transport state parameters in a normal state; s310, inputting transport state parameter vectors of all time windows in a current transport task of a financial client into a trained transport state parameter coupling analysis model, calculating a reconstruction error of the input vector by the model, taking the reconstruction error as a measure of deviation of nonlinear coupling relations among all transport state parameters in a corresponding time window from a normal state, and converting the corresponding reconstruction error into an abnormal risk probability value as a probability of occurrence of unauthorized path deviation abnormal risk in the corresponding time window; S311, obtaining a historical reference risk coefficient of a road grade corresponding to a route segment matched with each time window in the current transportation task of the financial client; S312, obtaining maximum values and minimum values in weighted anomaly scores of all time windows, taking a difference value between the maximum values and the minimum values as a weighted anomaly score range, dividing the difference value between the weighted anomaly scores of all the time windows and the minimum values by the weighted anomaly score range to obtain abnormal risks in the transportation process of all the time windows, and carrying out arithmetic average operation on the abnormal risks in the cargo state of all the time windows in the current transportation task of the financial client to obtain the abnormal risks in the cargo state of the current transportation task of the financial client.
  5. 5. The financial customer data management method based on big data evaluation according to claim 4, wherein the construction process of the risk fusion evaluation function preset in step S4 comprises the following specific steps: S41, extracting cargo payment triggering potential energy of current transported cargoes of financial clients and cargo state abnormal risks of current transportation tasks; S42, constructing a two-person zero-game model, taking the cargo pay triggering potential energy and the cargo state abnormal risk of the current transportation task as benefits of two game parties, solving hybrid strategy Nash equilibrium through an iterative algorithm, and obtaining the optimal weight w of the cargo pay triggering potential energy in fusion evaluation and the optimal weights 1-w of the cargo state abnormal risk of the current transportation task; S43, calculating the goods settlement triggering potential energy according to the settlement risk=w×the goods settlement triggering potential energy+ (1-w) x the abnormal risks of the goods state of the current transportation task, and obtaining the settlement risk goods settlement triggering potential energy of the current transportation task of the financial client.
  6. 6. The method for managing financial client data based on big data evaluation according to claim 5, wherein the step S5 of carrying out early warning of payment on goods of the financial client as the applicant based on the result of risk assessment of payment of the financial client on the current transportation task specifically comprises: S51, obtaining a pay risk assessment result of a current transportation task of a financial client; S52, presetting a claim risk threshold, carrying out goods claim early warning on the current transportation task of the financial client when the claim risk assessment result of the current transportation task of the financial client is larger than the claim risk threshold, carrying out emergency maintenance on goods associated with the current transportation task, and marking the current transportation task as completed transportation task when the claim risk assessment result of the current transportation task of the financial client is smaller than or equal to the claim risk threshold.
  7. 7. A big data evaluation based financial customer data management system implemented based on the big data evaluation based financial customer data management method of any of claims 1-6, the system comprising: The data acquisition module is used for acquiring the main body attribute data and the historical financial transaction data of the financial client serving as the insuring party, and simultaneously acquiring the goods transportation data and the transportation route data of the goods associated with the financial client serving as the insuring party; The system comprises a transport claim analysis module, a historical damage claim settlement data set, a standard function influence weight, a cargo claim payment triggering potential energy prediction model and a financial customer, wherein the transport claim analysis module is used for extracting a task number, a cargo type code, a damage position code, a damage area occupation ratio, an average depth change value of a damage position and a binary mark of whether claim settlement occurs to a corresponding transport task in the historical transport task record according to a completed historical transport task record in historical financial transaction data; The system comprises a cargo state detection module, a cargo state abnormal risk detection module, a cargo state analysis module and a cargo state analysis module, wherein the cargo state detection module is used for dividing the total duration of each transport task in a historical cargo state normal data set into a plurality of time windows respectively, extracting the position change parameter, the speed fluctuation parameter, the driving shock parameter, the environment deviation parameter, the sealing state parameter and the mechanical attenuation state risk coefficient representing the self mechanical property attenuation of the transport carrier from cargo transport data of each time window as transport state parameters representing each time window, training a cargo state abnormal risk detection model based on the nonlinear coupling relation of the transport state parameters of a multi-dimensional time sequence by utilizing the transport state parameters of each time window in the historical cargo state normal data set, judging the probability of unauthorized path deviation abnormal risk in the corresponding time window according to the comprehensive analysis of five transport state parameters, and analyzing the cargo state abnormal risk of the current transport task of a financial client based on the probability of unauthorized path deviation abnormal risk in the corresponding time window, and the historical risk coefficient matched with each time window in transport route data; the claim risk detection module is used for inputting the cargo claim triggering potential energy analysis result of the financial client and the cargo state abnormal risk analysis result of the current transportation task of the financial client into a preset risk fusion evaluation function based on a dynamic game theory, and evaluating the cargo claim triggering potential energy of the claim risk of the current transportation task of the financial client; And the claim early warning module is used for carrying out claim early warning on goods of the financial client serving as an applicant party based on the claim risk assessment result of the current transportation task of the financial client.

Description

Financial customer data management method and system based on big data evaluation Technical Field The invention relates to the technical field of financial client data management, in particular to a financial client data management method and system based on big data evaluation. Background In modern supply chain financial systems, cargo transportation insurance is a core financial tool for guaranteeing trade safety and dispersing logistics risks, and the accuracy and efficiency of the claim processing are directly related to the operation cost of insurance companies, customer satisfaction and even the stability and toughness of the whole supply chain. In the long-distance transportation process of cargoes, the cargoes can be damaged at the starting point or the end point due to static factors such as loading and unloading operation, inherent defects of packaging and the like, and can continuously bear dynamic risks in the complex transportation process, namely instantaneous impact force caused by sudden braking and jolting of a vehicle, erosion of the specific cargoes caused by severe fluctuation of ambient temperature and humidity, and abnormal risks of the cargoes possibly implied by abnormal stay or path deviation. These risk factors range from subtle potential damage to local functional failure until loss of the overall value of the cargo is initiated, which may lead to claims disputes and customer churning if their probability of occurrence and degree of loss are not assessed prospectively and intervened in time based on the data. However, the prior art simply predicts future risk by only paying attention to historical attributes of the financial customers themselves and transaction data, such as by only the frequency of claims, when constructing a risk management model for the goods transportation claims, without deeply correlating the microscopic decisions of customer claims with each specific, sensor-recorded detail of the goods damage and the functional importance of the damaged parts. This results in the model not being able to answer claims of similar value, the risk logic behind which is quite different, for example scraping of the container housing and cracking of the precision instrument spindle, although the maintenance costs may be similar, the latter causing a customer claim and resulting in a much higher probability of subsequent business interruption than the former, the risk being substantially different. Meanwhile, although the monitoring of the Internet of things is introduced in the prior art, stream data such as vibration, temperature and humidity in the transportation process are often analyzed in isolation, and the method is only used for backtracking whether a super-threshold event occurs afterwards, and the dynamic process data and inherent risks of a road network carrying transportation (such as the historical accident probability of a mountain area sharp bend section) are not subjected to spatial superposition analysis, and are not subjected to fusion research and judgment with static risk images of clients. Therefore, after the damage occurs, the damage is scientifically judged to be in a high-risk transportation process, so that the probability of final payment is difficult to accurately evaluate and the responsibility is reasonably determined, the risk management is stopped at a passive response level, and the active prediction and damage reduction capability is lacking. Disclosure of Invention In order to overcome the defects and shortcomings in the prior art, the invention provides a financial customer data management method and system based on big data evaluation, which respectively construct a claim risk analysis model based on logistic regression and an isolated forest process anomaly evaluation model fusing route risks by comprehensively acquiring customer static properties, historical transaction, cargo damage details, whole-course transportation sensing and route data, and perform weighted fusion on the outputs of the two models to generate quantitative evaluation on comprehensive claim risks of the current transportation task, thereby executing intelligent early warning. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, an embodiment of the present invention provides a financial customer data management method based on big data evaluation, including the steps of: s1, acquiring main body attribute data and historical financial transaction data of a financial customer serving as an applicant, and simultaneously acquiring cargo transportation data and transportation route data of related cargoes of the financial customer serving as the applicant; S2, extracting a task number, a cargo type code, a damage position code, a damage area occupation ratio, an average depth change value of a damage part and a binary mark of whether a corresponding transportation task generates a claim according to a completed