Financial Customer Data Management Methods and Systems Based on Big Data Assessment

Abstract

This invention relates to the field of financial customer data management technology, and in particular to a financial customer data management method and system based on big data assessment. This invention combines the financial customer's main attribute data and historical financial transaction data to analyze the potential for triggering claims for the financial customer's goods; it combines the cargo transportation data and transportation route data of the financial customer's associated goods to analyze the risk of abnormal cargo status in the current transportation task; based on the analysis results of the financial customer's potential for triggering claims for the financial customer's goods, and the analysis results of the risk of abnormal cargo status in the current transportation task, it assesses the claim risk of the financial customer's current transportation task; based on the assessment results of the claim risk of the financial customer's current transportation task, it provides a cargo claim warning; thereby improving the accuracy of risk pricing and claims review, effectively reducing potential cargo damage losses and claim costs, and enhancing customer service experience and risk management efficiency.

Description

Technical Field

[0001] This invention relates to the field of financial customer data management technology, and in particular to a financial customer data management method and system based on big data assessment. Background Technology

[0002] In a modern supply chain finance system, cargo transportation insurance is a core financial tool for ensuring trade security and mitigating logistics risks. The accuracy and efficiency of its claims processing directly affect the insurance company's operating costs, customer satisfaction, and even the stability and resilience of the entire supply chain. During long-distance transportation, goods are not only damaged at the origin or destination due to static factors such as loading and unloading operations and inherent packaging defects, but also continuously subjected to dynamic risks during the complex transportation journey: the instantaneous impact of sudden braking and bumps, the corrosive effects of drastic fluctuations in environmental temperature and humidity on specific goods, and the abnormal risks that may be implied by abnormal stops or route deviations. These risk factors range from causing minor potential damage to causing partial functional failure, and even leading to the loss of the overall value of the goods. If the probability of their occurrence and the extent of loss are not proactively assessed and timely intervened based on data, it may lead to claims disputes and customer loss.

[0003] However, existing technologies, when constructing risk management models for cargo transportation claims, only focus on the historical attributes and transaction data of financial clients. For example, they simply predict future risks based on claim frequency, without deeply linking the micro-level decisions of clients' claims with each specific detail of cargo damage recorded by sensors and the functional importance of the damaged components. This prevents the models from answering the core question of why claims of similar amounts have drastically different underlying risk logics. For instance, a scratch on a container shell and a crack in the spindle of a precision instrument may have similar repair costs, but the latter has a much higher probability of triggering a client claim and causing subsequent business interruption; their risks are fundamentally different. Furthermore, while existing technologies incorporate IoT monitoring, they often analyze flow data such as vibration, temperature, and humidity during transportation in isolation, only for retrospective analysis to determine if threshold events occurred. They fail to spatially overlay these dynamic process data with the inherent risks of the road network carrying the transportation (such as the historical accident probability on sharp bends in mountainous areas), nor do they integrate them with the static risk profile of the client. Therefore, it is impossible to scientifically determine the extent to which the damage originated from the high-risk transportation process after it occurs, making it difficult to accurately assess the probability of final compensation and reasonable liability determination. This leaves risk management at the level of passive response, lacking the ability to proactively foresee and mitigate losses. Summary of the Invention

[0004] To overcome the shortcomings and deficiencies of existing technologies, this invention provides a financial customer data management method and system based on big data assessment. By comprehensively acquiring customer static attributes, historical transactions, cargo damage details, and data from the entire transportation process, as well as route data, a claims risk analysis model based on logistic regression and an isolated forest process anomaly assessment model that integrates route risks are constructed respectively. The outputs of the two models are then weighted and fused to generate a quantitative assessment of the comprehensive compensation risk of the current transportation task, and intelligent early warning is executed accordingly.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a financial customer data management method based on big data assessment, comprising the following steps:

[0007] S1. Obtain the main attribute data and historical financial transaction data of the financial customer who is the insured party, and at the same time obtain the cargo transportation data and transportation route data of the goods associated with the financial customer who is the insured party.

[0008] S2. Based on the historical transportation task records completed in historical financial transaction data, extract the task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and binary indicators of whether a claim was made for the corresponding transportation task from the historical transportation task records; at the same time, extract the historical claim rate of the corresponding customer from the main attribute data, construct a historical damage claim dataset, and based on the historical damage claim dataset, statistically calculate the percentage of records with claims made under each pairwise combination of cargo type code and damage location code to the total number of records, define it as the baseline function influence weight corresponding to each pairwise combination of cargo type code and damage location code; thereby, through the historical damage claim dataset and the baseline function influence weight, construct a cargo claim trigger potential prediction model to analyze the cargo claim trigger potential of financial customers;

[0009] S3. Divide the total duration of each transportation task in the historical normal cargo status dataset into multiple time windows. For each time window, extract the position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical attenuation risk coefficient (characterizing the mechanical performance degradation of the transport vehicle) from the cargo transportation data as transportation status parameters for each time window. Using the transportation status parameters of each time window in the historical normal cargo status dataset, train a cargo status anomaly risk detection model based on the nonlinear coupling relationship of multi-dimensional time series transportation status parameters. This model is used to determine the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window based on the comprehensive analysis of the five transportation status parameters. Based on the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window output by the cargo status anomaly risk detection model and the historical benchmark risk coefficient matching each time window in the transportation route data, analyze the cargo status anomaly risk of the current transportation task for the financial customer.

[0010] Among them, the position change parameter is used to identify abnormal position changes and discover the risk of path deviation; the speed fluctuation parameter and driving aggressive parameter are used to identify driving behavior; the environmental deviation parameter is used to determine whether the cargo storage conditions are compliant; the sealing status parameter is used to detect whether the cargo may be illegally touched; and the mechanical deterioration status risk coefficient, which characterizes the mechanical performance degradation of the transport vehicle itself, is used to identify the potential transportation interruption or cargo loss risk caused by the deterioration of the technical condition of the transport vehicle.

[0011] S4. Input the analysis results of the potential energy for cargo compensation of financial clients and the analysis results of the abnormal cargo status risk of the current transportation task of financial clients into the preset risk fusion assessment function based on dynamic game theory to assess the compensation risk of the current transportation task of financial clients.

[0012] S5. Based on the risk assessment results of the current transportation task of the financial customer, issue a claim warning for the goods of the financial customer who is the insured party.

[0013] According to the above scheme, step S2 involves constructing a cargo claim trigger potential prediction model by using historical damage claims datasets and benchmark function influence weights to analyze the cargo claim trigger potential of financial clients; this includes the following specific steps:

[0014] S23. Based on the logistic regression model and the historical damage claims dataset, construct and train a cargo compensation trigger potential prediction model;

[0015] S24. For the goods currently being transported by the financial customer, obtain the damage location code, damage area percentage, and average depth change value of the damaged parts of the goods through image recognition and 3D scanning, and obtain the goods type code of the goods; based on the goods type code and damage location code, query the baseline function influence weight corresponding to the pairwise combination of the goods type code and damage location code, and obtain the historical claim rate of the corresponding customer; input the damage area percentage, average depth change value of the damaged parts, the corresponding baseline function influence weight, the historical claim rate of the corresponding customer, and the goods type code of the goods currently being transported into the trained goods claim trigger potential prediction model; the model outputs the trigger claim probability value of the goods currently being transported, and uses the trigger claim probability value of the goods currently being transported as the goods claim trigger potential value of the financial customer's goods currently being transported.

[0016] According to the above scheme, step S3, which extracts position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical degradation risk coefficient (characterizing the degradation of the transport vehicle's own mechanical performance) from the cargo transportation data for each time window, includes the following specific steps:

[0017] S31. From the cargo transportation data, obtain the vehicle latitude and longitude coordinate sequence and instantaneous speed sequence of the transport vehicle corresponding to the transportation task; the longitudinal and lateral acceleration event signal sequence exceeding the preset intensity threshold collected through the vehicle controller local area network bus; the temperature reading sequence and relative humidity reading sequence collected through the environmental sensors fixed inside the cargo compartment; the electronic seal status signal sequence collected through the electronic seal installed at the cargo unit door; and the raw diagnostic data stream reflecting the working status of each subsystem of the vehicle collected in real time through the vehicle self-diagnosis system interface, wherein the electronic seal status signal includes a locked status signal and an unlocked status signal; from the transportation route data, obtain the route segment sequence associated with the digital map road network, and the historical benchmark risk coefficient of the road level corresponding to each route segment, and dynamically update it after feature fusion based on the historical accident rate of the corresponding road segment;

[0018] S32. Divide the transportation task into multiple time windows with equal time intervals according to the duration of the transportation task, and obtain the cargo transportation data and transportation route data corresponding to each time window in the transportation task.

[0019] S33. Calculate the spherical linear displacement distance based on the vehicle's latitude and longitude coordinates at the beginning and end of the corresponding time window, and use it as the position change parameter for the corresponding time window;

[0020] S34. Calculate the standard deviation of all instantaneous velocity reading sequences within the corresponding time window, and use it as the velocity fluctuation parameter for the corresponding time window;

[0021] S35. The total number of longitudinal and lateral acceleration event signals triggered within the corresponding time window is accumulated and used as the driving aggressive parameter for the corresponding time window;

[0022] S36. Determine whether the reading in any of the temperature reading sequence and relative humidity reading sequence within the corresponding time window exceeds the preset safety threshold range. If it does, set the environmental deviation parameter of the corresponding time window to 1; otherwise, set the environmental deviation parameter of the corresponding time window to 0.

[0023] S37. Identify whether the electronic seal status signal sequence within the corresponding time window changes from a locked status signal to an unlocked status signal. If it does, assign the sealing status parameter of the corresponding time window to 1; otherwise, assign it to 0.

[0024] S38. Extract features from the raw diagnostic data stream reflecting the working status of each vehicle subsystem collected within the corresponding time window. Calculate the degree of mechanical performance degradation of the transport vehicle within the current time window based on the extracted feature values, using this as a mechanical performance degradation risk coefficient characterizing the degradation of the transport vehicle's own mechanical performance. Step S3 involves analyzing the abnormal cargo status risk of the financial customer's current transport task, including the following specific steps:

[0025] S39. Construct a transportation state parameter coupling analysis model based on deep neural network. Take the transportation state parameter vector of each time window in the historical cargo state normal dataset as input and train it with the goal of reconstructing the corresponding vector. This enables the model to learn the nonlinear coupling relationship between various transportation state parameters under normal conditions.

[0026] S310. Input the transportation status parameter vectors of all time windows in the current transportation task of the financial customer into the trained transportation status parameter coupling analysis model, calculate the reconstruction error of the model to the input vector, use the reconstruction error as a measure of the deviation of the nonlinear coupling relationship between each transportation status parameter in the corresponding time window from the normal state, and convert the corresponding reconstruction error into an abnormal risk probability value as the probability of an abnormal risk of unauthorized path deviation occurring in the corresponding time window.

[0027] S311. Obtain the historical baseline risk coefficient of the road level corresponding to the route segment that matches each time window in the current transportation task of the financial customer; multiply the abnormal risk probability value corresponding to each time window in the current transportation task of the financial customer by the historical baseline risk coefficient of the corresponding road level to obtain the weighted abnormal score of each time window.

[0028] S312. Obtain the maximum and minimum values ​​of the weighted anomaly scores for all time windows; use the difference between the maximum and minimum values ​​as the weighted anomaly score range; divide the difference between the weighted anomaly score and the minimum value for each time window by the weighted anomaly score range to obtain the transportation process anomaly risk for each time window; perform an arithmetic average of the cargo status anomaly risks for all time windows in the current transportation task of the financial customer to obtain the cargo status anomaly risk for the current transportation task of the financial customer.

[0029] According to the above scheme, the construction process of the preset risk fusion assessment function in step S4 includes the following specific steps:

[0030] S41. Extract the potential energy for cargo compensation triggering the current transportation of goods by financial clients and the abnormal cargo status risk of the current transportation task;

[0031] S42. Construct a two-player zero-sum game model, taking the cargo compensation triggering potential and the cargo state abnormality risk of the current transportation task as the payoffs of the two players; solve the mixed strategy Nash equilibrium through an iterative algorithm to obtain the optimal weight w of the cargo compensation triggering potential in the fusion evaluation and the optimal weight 1-w of the cargo state abnormality risk of the current transportation task.

[0032] S43. Based on the formula: compensation risk = w × cargo compensation trigger potential energy + (1-w) × cargo status abnormality risk of the current transportation task, calculate the compensation risk and cargo compensation trigger potential energy of the current transportation task for the financial customer.

[0033] According to the above scheme, step S5, based on the risk assessment results of the financial customer's current transportation task, provides a risk warning for the financial customer's goods as the insured party, specifically including:

[0034] S51. Obtain the risk assessment results of the compensation for the current transportation task of the financial customer;

[0035] S52. Preset compensation risk threshold: When the compensation risk assessment result of the financial customer's current transportation task is greater than the compensation risk threshold, issue a cargo compensation warning for the financial customer's current transportation task and perform emergency maintenance on the cargo associated with the current transportation task; when the compensation risk assessment result of the financial customer's current transportation task is less than or equal to the compensation risk threshold, mark the current transportation task as a completed transportation task.

[0036] Secondly, embodiments of the present invention also provide a financial customer data management system based on big data assessment, including:

[0037] The system includes:

[0038] The data acquisition module is used to acquire the main attribute data and historical financial transaction data of the financial customer who is the insured party, and at the same time acquire the cargo transportation data and transportation route data of the goods associated with the financial customer who is the insured party.

[0039] The transportation claims analysis module is used to extract the task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and a binary indicator of whether a claim has been made for the corresponding transportation task from historical transportation task records completed in historical financial transaction data. Simultaneously, it extracts the historical claim rate for the corresponding customer from the main attribute data, constructs a historical damage claims dataset, and statistically calculates the percentage of records with claims made under each pairwise combination of cargo type code and damage location code, defining this as the baseline functional influence weight for each pairwise combination of cargo type code and damage location code. Therefore, through the historical damage claims dataset and the baseline functional influence weight, a cargo claim triggering potential prediction model is constructed to analyze the cargo claim triggering potential of financial customers.

[0040] The cargo status detection module divides the total duration of each transportation task in the historical normal cargo status dataset into multiple time windows. For each time window, it extracts position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical attenuation risk coefficients representing the mechanical performance degradation of the transport vehicle from the cargo transportation data. These parameters characterize the transportation status of each time window. Using these transportation status parameters from the historical normal cargo status dataset, a cargo status anomaly risk detection model based on the nonlinear coupling relationship of multidimensional time series transportation status parameters is trained. This model is used to determine the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window based on a comprehensive analysis of the five transportation status parameters. Based on the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window output by the cargo status anomaly risk detection model and the historical benchmark risk coefficients matching each time window in the transportation route data, the module analyzes the cargo status anomaly risk of the current transportation task for the financial client.

[0041] The claims risk detection module is used to input the analysis results of the cargo claims triggering potential of financial customers and the analysis results of the abnormal cargo status risk of the current transportation task of financial customers into a preset risk fusion assessment function based on dynamic game theory, and to assess the claims risk of the current transportation task of financial customers and the cargo claims triggering potential.

[0042] The claims warning module is used to issue claims warnings for the goods of financial clients who are the insured parties, based on the claims risk assessment results of the current transportation task.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] 1. This invention achieves refined and objective prediction of claims risks by quantitatively analyzing the correlation between the location and extent of damage to goods and their functional importance, and by combining this with the customer's historical behavior. This overcomes the assessment bias caused by the reliance on qualitative experience in traditional methods, and improves the accuracy of risk pricing and claims review.

[0045] 2. This invention integrates multi-source IoT data with historical risk knowledge of road networks to construct a dynamic process anomaly risk scoring model, which upgrades the risk assessment during transportation from discrete event alarms to continuous situation assessment, providing quantifiable decision-making basis for process risk management.

[0046] 3. This invention intelligently integrates and assesses static damage risk and dynamic process risk, and provides early warning based on this, effectively reducing potential cargo damage losses and compensation costs, and improving customer service experience and risk management efficiency. Attached Figure Description

[0047] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 This is a schematic diagram of the overall process of the financial customer data management method based on big data assessment of the present invention;

[0049] Figure 2 This is a flowchart of step S2 in the financial customer data management method based on big data assessment of the present invention.

[0050] Figure 3 This is a schematic diagram of the structure of the financial customer data management system based on big data assessment according to the present invention;

[0051] Figure 4 This is a flowchart illustrating the workflow of the cargo status anomaly risk detection model in the financial customer data management method based on big data assessment of the present invention. Detailed Implementation

[0052] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0053] Example 1

[0054] like Figure 1 , Figure 2 and Figure 4 As shown, this embodiment provides a financial customer data management method based on big data assessment, which specifically includes the following steps:

[0055] S1. Obtain the main attribute data and historical financial transaction data of the financial customer who is the insured party, and at the same time obtain the cargo transportation data and transportation route data of the goods associated with the financial customer who is the insured party.

[0056] S2. Based on the historical transportation task records completed in historical financial transaction data, extract the task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and binary indicators of whether a claim was made for the corresponding transportation task from the historical transportation task records; at the same time, extract the historical claim rate of the corresponding customer from the main attribute data, construct a historical damage claim dataset, and based on the historical damage claim dataset, statistically calculate the percentage of records with claims made under each pairwise combination of cargo type code and damage location code to the total number of records, define it as the baseline function influence weight corresponding to each pairwise combination of cargo type code and damage location code; thereby, through the historical damage claim dataset and the baseline function influence weight, construct a cargo claim trigger potential prediction model to analyze the cargo claim trigger potential of financial customers;

[0057] S3. Divide the total duration of each transportation task in the historical normal cargo status dataset into multiple time windows. For each time window, extract the position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical attenuation risk coefficient (characterizing the mechanical performance degradation of the transport vehicle) from the cargo transportation data as transportation status parameters for each time window. Using the transportation status parameters of each time window in the historical normal cargo status dataset, train a cargo status anomaly risk detection model based on the nonlinear coupling relationship of multi-dimensional time series transportation status parameters. This model is used to determine the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window based on the comprehensive analysis of the five transportation status parameters. Based on the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window output by the cargo status anomaly risk detection model and the historical benchmark risk coefficient matching each time window in the transportation route data, analyze the cargo status anomaly risk of the current transportation task for the financial customer.

[0058] Among them, the position change parameter is used to identify abnormal position changes and discover the risk of path deviation; the speed fluctuation parameter and driving aggressive parameter are used to identify driving behavior; the environmental deviation parameter is used to determine whether the cargo storage conditions are compliant; the sealing status parameter is used to detect whether the cargo may be illegally touched; and the mechanical deterioration status risk coefficient, which characterizes the mechanical performance degradation of the transport vehicle itself, is used to identify the potential transportation interruption or cargo loss risk caused by the deterioration of the technical condition of the transport vehicle.

[0059] S4. Input the analysis results of the potential energy for cargo compensation of financial clients and the analysis results of the abnormal cargo status risk of the current transportation task of financial clients into the preset risk fusion assessment function based on dynamic game theory to assess the compensation risk of the current transportation task of financial clients.

[0060] S5. Based on the risk assessment results of the current transportation task of the financial customer, issue a claim warning for the goods of the financial customer who is the insured party.

[0061] In this embodiment, as Figure 2 As shown, step S2 extracts the historical claim rate of the corresponding customer from the main attribute data, constructs a historical damage claim dataset, and based on the historical damage claim dataset, statistically calculates the percentage of records with claims occurring under each pairwise combination of cargo type code and damage location code to the total number of records, which is defined as the baseline function influence weight corresponding to each pairwise combination of cargo type code and damage location code; including the following specific steps:

[0062] S21. Extract all completed historical transportation task records from historical financial transaction data. These records include: task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and a binary indicator indicating whether a claim was made for the corresponding transportation task. The damage area percentage is the percentage of the damaged area relative to the total surface area of ​​the component at that location, calculated by identifying the damaged area image using an image analysis system. The average depth change of the damaged area is the absolute value of the average depth of deformation calculated by scanning the damaged area with a 3D scanning device and comparing it with a standard 3D model. Extract the historical claim rate for the corresponding customer from the financial customer's subject attribute data. The historical claim rate is the percentage of transportation tasks for which claims were made out of the total number of transportation tasks in the customer's history. Use all completed historical transportation task records and the corresponding customer's historical claim rate as the historical damage claim dataset. It should be noted that this step is the data foundation for constructing a quantitative analysis of claim risk. Its core function is to transform the relatively static and abstract transaction records in traditional financial scenarios into structured data that is deeply bound to the physical state of the goods and can be quantified and measured. In conventional financial customer data management, such as in credit approval scenarios, data analysis mainly revolves around the customer's financial and credit dimensions, with data attributes leaning towards abstract value symbols and credit history. However, in the specific scenario of cargo transportation claims, financial risk is directly rooted in physical damage in the physical world. Therefore, data management must extend to the precise perception and description of the goods. Step S21 introduces two objective parameters directly measured by instruments (image analysis system, 3D scanning equipment): the proportion of damaged area and the average depth change value of damaged parts. This achieves a leap from qualitative description to quantitative analysis of cargo damage. This differs from other scenarios that rely on subjective assessments or qualitative reports. For example, in assessing customer risk preferences for wealth management products, the data may come from customer-filled questionnaires, which are subjective; while the damage parameters in this step are objective and repeatable physical quantities. By extracting cargo type codes and damage location codes, this embodiment standardizes and classifies the diverse cargo and its components, making subsequent statistical analysis and large-scale machine learning possible. Associating the binary marker of whether a claim has occurred with these refined objective damage parameters is key to establishing the feature-label correspondence for subsequent modeling. At the same time, by introducing the customer's historical claims rate, a dimension derived from traditional financial data management, we can combine the customer's long-term credit behavior with the objective fact of a single cargo damage. This is a manifestation of the multidimensionality of data management in the cargo transportation claims scenario—it is necessary to pay attention to both the objective condition of the goods and the historical behavior patterns of the people (customers).Based on the above, the entire dataset construction process in this embodiment essentially involves associating and merging fragmented related data through a unified task number to form a high-quality data cube that can be used for in-depth analysis, laying a solid foundation for mining claims logic from the data.

[0063] S22. Based on historical damage claims datasets, statistically calculate the percentage of claims records occurring under each pairwise combination of cargo type code and damage location code, out of the total number of records. This percentage is defined as the baseline functional impact weight corresponding to each pairwise combination of cargo type code and damage location code. It should be noted that the core function of this step is to automatically learn and quantify the functional importance of different damages for different goods from historical big data, thereby transforming qualitative engineering experience into quantitative risk coefficients. This is a crucial step in achieving intelligent claims decision-making. This step focuses on specific, physical combinations of cargo type and damage location, rather than abstract financial transaction behaviors. The detailed process for obtaining the baseline functional impact weight is as follows: First, read the complete historical damage claims dataset, ensuring that the data covers a sufficiently long time period and a large enough sample size to guarantee the stability of the statistical results. Next, group the data according to the two key fields of cargo type code and damage location code. For example, group all historical records where the cargo type code is precision machine tool (let's assume code A01) and the damage location code is spindle box (let's assume code P05). Within each such group, perform two statistical operations: First, calculate the total number of historical records in the group. Second, iterate through each record in the group and check its binary flag field indicating whether a claim has occurred, counting the number of records marked as "yes" (i.e., a claim has occurred). Finally, use the proportion of records marked as "yes" (i.e., a claim has occurred) to the total number of historical records as the baseline functional impact weight for that group. Specifically, this means that when cargo of type A01 is damaged at location P05 in the past, the probability of ultimately resulting in a claim is reflected. This value objectively reflects the degree of impact of damage at this location on the functional integrity of this type of cargo. For example, for precision machine tools, the weight of spindle box damage may be as high as 0.90 (meaning that customers will claim compensation in 90% of cases), while the weight of paint scratches on the casing may be only 0.1. Based on the above, such a weight is calculated for all combinations of goods type and damage location that have appeared in historical data, forming a continuously updated weight dictionary.

[0064] In this embodiment, step S2 constructs a cargo claim trigger potential prediction model by using historical damage claims datasets and benchmark function influence weights to analyze the cargo claim trigger potential of financial customers; this includes the following specific steps:

[0065] S23. Based on the logistic regression model and historical damage claims dataset, construct and train a cargo claim triggering potential prediction model. Use the percentage of damaged area, average depth change of damaged parts, corresponding baseline function influence weight, historical claim rate of the corresponding customer, and cargo type code of each historical transportation task record in the historical damage claims dataset as input features of the model. Use the binary label indicating whether a claim has occurred for the corresponding transportation task as the output target label of the model. Randomly divide the historical damage claims dataset into a model training set and a model validation set according to a 7:3 ratio of historical transportation task records. Use the model training set to train the logistic regression model to obtain the initial prediction model. Validate the initial prediction model using the model validation set. The initial prediction model whose accuracy on the validation set reaches the preset accuracy is used as the final cargo claim triggering potential prediction model. It should be noted that this step is the core link in transforming the previously processed structured data into usable predictive intelligence. Its role is to establish a quantitative prediction model that can comprehensively consider the objective severity of damage, the importance of damaged components, and the customer's claim tendency. Unlike models used in credit scenarios to predict customer default risk, this model predicts the probability of triggering a financial compensation (claims) event under specific physical damage. Its input features combine physical world measurements and historical financial behavior data, reflecting the integration of scenarios. The process for obtaining model input parameters has been detailed in the preceding steps: the proportion of damage area and the average depth change of the damage site are obtained from instrument measurements; the baseline function influence weights are derived from historical statistics of S22; the customer's historical claims rate is calculated from customer attribute data; and the cargo type code is a classification identifier. These features require necessary preprocessing before being input into the model. For example, for categorical variables such as cargo type codes, this embodiment can further employ techniques such as one-hot encoding to convert them into numerical features; for numerical features, such as the proportion of damage area, this embodiment can further standardize them as needed to eliminate the influence of dimensions and ensure the stability of model training. The logistic regression model was chosen in this embodiment because of its good interpretability. The feature coefficients generated after model training can intuitively reflect the direction and strength of each factor's influence on the claim probability, which meets the requirements of transparency and auditability of risk models in the financial field. The specific process of model training is as follows: 70% of the historical data (training set) is input into the logistic regression algorithm. The algorithm iteratively adjusts the coefficients corresponding to each input feature through optimization methods such as maximizing the likelihood function, with the goal of minimizing the difference between the model's predicted claim probability and the actual claim status. After training, the remaining 30% of the data (validation set) is used to perform an unbiased evaluation of the model's performance. The overall accuracy is calculated during the validation process.The preset accuracy is a business decision value. For example, in this embodiment, the preset accuracy can be set to 85%, meaning that the model will only be deployed if its prediction accuracy on the validation set reaches or exceeds 85%. If it does not reach 85%, it is necessary to backtrack and check data quality, feature engineering, or adjust model parameters. This rigorous training-validation process ensures the model's generalization ability and reliability, avoiding overfitting to the training data. The final retained model is a mathematical function that can map the multi-dimensional data of a cargo damage event (how much, how deep, where, who the customer is) to a claim probability. This probability value is the cargo compensation triggering potential in subsequent steps; thus, a reliable quantitative conversion from physical damage to financial risk is achieved.

[0066] S24. For the goods currently being transported by the financial customer, obtain the damage location code, damage area percentage, and average depth change value of the damaged parts through image recognition and 3D scanning, and obtain the cargo type code of the current transported goods; based on the cargo type code and damage location code, query the corresponding baseline function influence weights for each pair of cargo type code and damage location code, and obtain the historical claims rate of the corresponding customer; input the damage area percentage, average depth change value of the damaged parts, corresponding baseline function influence weights, historical claims rate of the corresponding customer, and cargo type code of the current transported goods into the trained cargo claim triggering potential prediction model; the model outputs the claim occurrence probability value of the current transported goods, and uses the claim occurrence probability value of the current transported goods as the cargo claim triggering potential value of the financial customer's current transported goods. It should be noted that this step is a key operational step in applying the aforementioned constructed static data capabilities and prediction model to dynamic business scenarios. Its role is to conduct real-time and automated claims risk rating for a single, specific cargo damage event. This is different from the periodic customer credit review in other financial scenarios. This step is an immediate risk assessment triggered by the specific event of cargo damage. When damage is discovered to goods upon completion of a transport mission, surveyors can use specialized equipment (such as high-resolution cameras and handheld 3D scanners) to collect data on the damaged areas. The image analysis system automatically analyzes the photographs, identifies the damaged area through image segmentation algorithms, and determines the damaged component based on a pre-set cargo component template (associated with the damage location code), thereby calculating the percentage of the damaged area. The 3D scanning equipment acquires point cloud data of the damaged area, compares it with a standard, intact 3D digital model, and calculates the average depth change value of the dent or deformation. Simultaneously, the cargo type code for this batch of goods is obtained from the transport documents. These objective measurement data, along with the baseline functional influence weight of the cargo type-damage location combination obtained through real-time queries and the customer's historical claims rate, constitute a complete and standardized set of model input feature vectors. This set of vectors is input into the S23-trained, deployed cargo claims trigger potential prediction model. The model performs a forward calculation, comprehensively weighing the physical severity of the damage, its inherent importance, and the customer's claim habits based on its inherent logical relationships learned from historical data. The final output is a continuous value between 0 and 1, representing the probability of a claim. For example, an output of 0.85 indicates that the model predicts an 85% chance that the damage will lead to a valid claim. This probability value is the quantified potential for triggering cargo compensation. It is no longer a qualitative assessment of high or low risk, but a value that can be precisely compared with risk thresholds in subsequent steps. This step significantly improves the efficiency and scientific rigor of the front-end claims processing, transforming the damage assessment and initial risk judgment work, which previously relied on the surveyor's personal experience, into a data-driven, standardized, and automated process.This not only reduces human error, but also provides a precise basis for decision-making in subsequent differentiated processing procedures (such as fast claims processing channels or enhanced review channels).

[0067] In this embodiment, step S3 divides the transportation task into multiple time windows, and extracts the position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, and sealing status parameters for each time window from the cargo transportation data, including the following specific steps:

[0068] S31. From the freight transportation data, obtain the vehicle's latitude and longitude coordinate sequence and instantaneous speed sequence corresponding to the transportation task; the longitudinal and lateral acceleration event signal sequence exceeding the preset intensity threshold collected through the vehicle controller local area network bus; the temperature reading sequence and relative humidity reading sequence collected through environmental sensors fixed inside the cargo compartment; and the electronic seal status signal sequence collected through the electronic seal installed at the freight unit door, where the electronic seal status signal includes locked status signal and unlocked status signal; from the transportation route data, obtain the route segment sequence associated with the digital map road network, and the historical baseline risk coefficient of the road level corresponding to each route segment, and dynamically update it after feature fusion based on the historical accident rate of the corresponding road segment; it should be noted that this step is the data collection and aggregation stage for implementing transportation process risk analysis, and its core role is to construct a multi-dimensional, time-series data set covering the behavior of the transportation vehicle, the micro-environment of freight transportation, and the context of the planned route. The risk of freight transportation is a continuous process that accumulates in physical space over time, therefore, this step collects high-frequency spatiotemporal trajectory and sensor time-series data, reflecting the management of dynamic processes. Specifically, the vehicle's latitude and longitude coordinate sequence and instantaneous velocity sequence are typically generated by the vehicle's GPS device at a frequency of once per second or higher, and transmitted to the data platform in real time or near real time via wireless communication networks (such as 4G / 5G), forming a spatiotemporal line describing the vehicle's macroscopic motion state; the longitudinal and lateral acceleration event signal sequences originate from the vehicle's CAN bus. When the inertial force of acceleration, braking, or turning exceeds a preset physical threshold (e.g., longitudinal acceleration greater than 0.4g or less than -0.5g, lateral acceleration greater than 0.3g), a timestamped event record is generated. These events are the direct causes of physical risks such as cargo displacement, collision, and tipping; the temperature and relative humidity reading sequences are periodically collected (e.g., once per minute) by IoT sensors deployed in the cargo compartment to monitor the transportation environment of temperature and humidity-sensitive goods such as cold chain pharmaceuticals and precision instruments; the electronic seal status signal sequence records every legal planned opening of the cargo compartment door (such as at the loading port and destination port) and every illegal abnormal opening, serving as important digital evidence of cargo safety and integrity. All these time-series data streams are aligned and correlated using unified transportation task numbers and standardized timestamps. On the other hand, transportation route data is pre-planned static information, including geographic polygon or linear reference information of each road segment (section) traversed from the origin to the destination, as well as the attributes of each road segment, such as road level (expressway, national highway, secondary mountain road, urban congestion section, etc.). Furthermore, in this embodiment, the historical accident rate of a corresponding road segment is defined as the ratio of the number of historical accidents of the corresponding road segment to the total number of historical accidents of all road segments. This step, by acquiring and preparing these two types of data, lays the foundation for subsequent analysis under unified spatiotemporal framework of what kind of abnormal events occurred under what road conditions.This is crucial data preparation for shifting from simply checking locations after the fact to perceiving risks before and during the event.

[0069] S32. Divide the transportation task into multiple time windows with equal time intervals according to the duration of the transportation task, and obtain the cargo transportation data and transportation route data corresponding to each time window in the transportation task.

[0070] S33. Based on the vehicle's latitude and longitude coordinates at the beginning and end of the corresponding time window, calculate the spherical linear displacement distance as the position change parameter for the corresponding time window. It should be noted that this step aims to quantify the vehicle's basic motion within a unit of time and extract the position change parameter, which is the basis for assessing the activity level and abnormal stillness during transportation. The detailed acquisition process is as follows: For each pre-divided time window, accurately extract two latitude and longitude coordinate points from the GPS data sequence of that window: the start time (e.g., second 0) and the end time (e.g., second 60). Since the Earth is a sphere, the Euclidean distance formula in a Cartesian coordinate system cannot be used; therefore, this embodiment uses a spherical distance calculation formula, such as the Haversine formula. This formula takes the latitude and longitude of two points as input and calculates the great circle distance between the two points along the Earth's surface through trigonometric operations; the specific calculation process is existing technology and will not be elaborated here.

[0071] S34. Calculate the standard deviation of all instantaneous speed reading sequences within the corresponding time window as the speed fluctuation parameter for that time window. The significance of using the speed fluctuation parameter in this embodiment is that: the larger the standard deviation, the more drastic the speed change within the time window, indicating that the vehicle may have experienced frequent acceleration and deceleration, which is usually related to congested road conditions, driving habits, or complex road environments. This unstable driving state directly increases the risk of cargo shifting or colliding due to inertial forces. Conversely, the smaller the standard deviation, the more stable the speed is, indicating smooth driving and relatively low risk. By calculating the speed standard deviation for each time window, this embodiment transforms the continuous speed curve into a series of discrete values ​​characterizing the intensity of fluctuations, thereby using statistical methods to identify high-risk periods of abnormal speed fluctuations.

[0072] S35. Accumulate the total number of longitudinal and lateral acceleration event signals triggered within the corresponding time window, using this as the driving aggressive parameter for the corresponding time window. It should be noted that this step aims to directly quantify the frequency of events that pose a direct physical impact on cargo safety during driving, and extract the driving aggressive parameter. This is the most direct link between vehicle dynamics and the physical causes of cargo damage. The acquisition process is detailed as follows: Two physical thresholds are pre-set to filter out risky acceleration events from the massive data on the CAN bus: a longitudinal acceleration threshold (e.g., +0.4g for rapid acceleration, -0.5g for emergency braking) and a lateral acceleration threshold (e.g., ±0.3g for sharp turns). When the vehicle's inertial measurement unit detects that the actual acceleration value exceeds these thresholds, the vehicle gateway generates an event message with a precise timestamp. For each time window, this embodiment retrieves all acceleration event records whose timestamps fall within the window's range; then, the number of longitudinal acceleration events (including rapid acceleration and emergency braking) and the number of lateral acceleration events are accumulated respectively. Finally, these two accumulated values ​​are added together to obtain the total number of aggressive driving events within the time window. This total number is the aggressive driving parameter for that window. For example, if there are two sudden braking events and one sharp turn within one minute, the parameter value is 3. This parameter has a very clear physical meaning: each event exceeding the threshold represents a significant inertial force impact on the cargo; frequent event accumulation greatly increases the probability of packaging damage, cargo displacement, and internal structural damage.

[0073] S36. Determine whether any reading in the temperature and relative humidity reading sequences within the corresponding time window exceeds a preset safety threshold range. If it does, assign a value of 1 to the environmental deviation parameter for the corresponding time window; otherwise, assign a value of 0. It should be noted that this step aims to monitor whether the microenvironment of cargo transportation is within an acceptable range and extract a binary environmental deviation parameter. The detailed acquisition process is as follows: Based on the most general requirements of the transported goods, preset a set of conservative temperature and humidity safety thresholds. For example, for most general cargo, the temperature range might be set to -5 to 35 degrees Celsius, and the relative humidity to 20% to 80%. This range should be broad enough to avoid generating too many alarms under normal climate changes, but must still cover basic protection against extreme environments. For each time window, extract all temperature and relative humidity readings within that window. These readings may be one or more per minute. Then, perform two independent judgments: First, check if any temperature reading within the window is below -5 degrees Celsius or above 35 degrees Celsius; second, check if any relative humidity reading within the window is below 20% or above 80%. If either of the above two judgments is true, it means that at least at one moment within this time window, the cargo environment exceeded the general safety boundary, and the environmental deviation parameter for that window is assigned a value of 1. Only when all temperature and humidity readings within the window strictly fall within the preset upper and lower limits is the parameter assigned a value of 0. The significance of using a binary parameter in this embodiment is that it is a clear risk indicator. A value of 1 indicates that the goods have been exposed to potentially harmful environmental conditions, which is particularly important for goods such as food, pharmaceuticals, electronic products, and paper. Even without visible physical damage, this could lead to hidden losses such as spoilage or failure. A parameter value of 0 indicates that the environment is normal within the monitoring range.

[0074] S37. Identify whether the electronic seal status signal sequence within the corresponding time window transitions from a locked state signal to an unlocked state signal. If so, assign a value of 1 to the sealing status parameter for the corresponding time window; otherwise, assign a value of 0. It should be noted that this step aims to monitor whether the physical integrity of the freight unit (container, cargo compartment) has been unplannedly interrupted, extracting the binary sealing status parameter; this is a core indicator for cargo security and theft prevention. The detailed acquisition process is as follows: After loading is completed and the compartment door is closed, the electronic seal will trigger a locking command remotely or on-site from the management platform, changing the seal status to locked. Subsequently, during transportation, any unauthorized attempt to open the door will result in the seal being damaged or triggering an unlocking signal, which will be reported via the wireless network. For each time window, extract the electronic seal status signal sequence arranged chronologically within that window. Then, check adjacent signal records one by one; the core logic is to find the transition point where the status changes from locked (LOCKED) to unlocked (UNLOCKED). If at least one such transition is detected in the signal sequence within a certain time window (e.g., one record is locked with a timestamp of 10:00:30, and the next record is unlocked with a timestamp of 10:00:45, and both timestamps fall within the same time window), then an unauthorized door opening event is determined to have occurred within that time window; subsequently, the seal status parameter for that window is assigned a value of 1. If no state transition occurs within the window, or the status remains unlocked (this should not occur during transit), then the parameter is assigned a value of 0. It is important to emphasize that when assigning a value to the seal status parameter, planned unlocking events must be excluded. For example, at the destination port, when a vehicle arrives at a designated geofenced area and is authorized to unlock, such events will have an authorization flag and should be filtered out when calculating this parameter. Therefore, this parameter captures unexpected, transshipment-related breaches of integrity. A parameter value of 1 directly indicates a potential breach of cargo. A parameter value of 0 indicates that the cargo is sealed intact during the monitoring period. The acquisition of this parameter relies on the reliable status reporting and accurate timestamps of the IoT seal, which can be achieved through a simple status transition detection logic, adding a crucial security dimension to the entire risk assessment system;

[0075] S38. Extract features from the original diagnostic data streams that reflect the working status of each subsystem of the vehicle collected within the corresponding time window. Calculate the degree of mechanical performance degradation of the transport vehicle within the current time window based on the extracted feature values. Use this as a mechanical performance degradation risk coefficient to characterize the mechanical performance degradation of the transport vehicle itself. It should be noted that the degree of mechanical performance degradation is determined through operational data analysis. In a specific embodiment, the average standard deviation of the operational data and the safe operational data over a period of time is used as the degree of mechanical performance degradation.

[0076] In this embodiment, step S3 analyzes the risk of abnormal cargo status for the current transportation task of the financial customer, including the following specific steps:

[0077] S39. Construct a transportation state parameter coupling analysis model based on deep neural network. Take the transportation state parameter vector of each time window in the historical cargo state normal dataset as input and train it with the goal of reconstructing the corresponding vector. This enables the model to learn the nonlinear coupling relationship between various transportation state parameters under normal conditions.

[0078] In one specific embodiment, the construction of the transportation status parameter coupling analysis model includes the following specific steps: Data source: using a historical cargo status normal dataset, which means that all data used for training must come from transportation tasks that ultimately deliver the cargo safely without any claims or abnormal reports;

[0079] Data format: Each data sample is a vector of transportation state parameters for a time window. That is, after dividing the time window according to S32, S33-S38 are executed for each window to generate a vector containing 6 values: [position change, speed fluctuation, aggressive driving, environmental deviation, sealing status, and mechanical decay]. Thousands of time window vectors for normal tasks constitute the training set of the model.

[0080] Model architecture design:

[0081] Input layer: The number of neurons is fixed at 6, corresponding to 6 transport state parameters;

[0082] Encoder section:

[0083] The first hidden layer may be designed with 10 neurons, using activation functions such as ReLU to introduce non-linearity.

[0084] Second hidden layer: further compressed, for example, designed to have 5 neurons;

[0085] Bottleneck layer (latent space): This is the end of encoding and the beginning of decoding. It is designed with a lower number of neurons than the input dimension, such as 3 neurons; this is key to the model learning the essence of normal patterns.

[0086] Decoder section (symmetric layer-by-layer reconstruction): Third hidden layer: symmetrical to the second hidden layer, for example, 5 neurons;

[0087] Fourth hidden layer: symmetrical to the first hidden layer, for example, 10 neurons;

[0088] Output layer: The number of neurons must be 6, the same as the input layer, corresponding to the 6 parameter values ​​to be reconstructed. A linear activation function is typically used.

[0089] Model training process:

[0090] Loss function definition: Using mean squared error, calculate the difference between the reconstructed vector output by the model and the original input vector at each parameter, square the difference, and then average the results. This value is the reconstruction error. The goal of training is to minimize the average reconstruction error of all training samples.

[0091] Training process:

[0092] Initialize all connection weights in the model to random small values;

[0093] Input normal parameter vectors into the model;

[0094] The model is forward-propagated to obtain the reconstructed output vector;

[0095] Calculate the reconstruction error (loss) of this sample;

[0096] The loss is propagated back from the output layer to the input layer layer by layer using the backpropagation algorithm, and the contribution of each weight to the error is calculated.

[0097] The optimizer is used to fine-tune all the weight values ​​in the model based on the calculated contribution, so that the error is reduced.

[0098] Repeat the above steps, iterating through all training data multiple times, until the total reconstruction error of the model no longer decreases significantly and tends to stabilize;

[0099] Training process parameters:

[0100] Learning rate: controls the step size of each parameter update; recommended value: 1e-3 to 3e-4; Batch size: how many samples are input in each iteration (e.g., 32, 64); affects training stability and speed; Number of epochs: how many rounds of training are traversed; recommended value: 100-200; Optimizer: selects the algorithm for parameter updates (e.g., Adam); Activation function: the non-linear function of neurons in the network, such as ReLU, used to learn complex patterns; Loss function: mean squared error;

[0101] Output the coupled analysis model of transportation state parameters obtained from training;

[0102] S310. Input the transportation status parameter vectors of all time windows in the current transportation task of the financial customer into the trained transportation status parameter coupling analysis model, calculate the reconstruction error of the model to the input vector, use the reconstruction error as a measure of the deviation of the nonlinear coupling relationship between each transportation status parameter in the corresponding time window from the normal state, and convert the corresponding reconstruction error into an abnormal risk probability value as the probability of an abnormal risk of unauthorized path deviation occurring in the corresponding time window.

[0103] The specific steps for converting the corresponding reconstruction error into anomaly risk probability values ​​are as follows: collect the reconstruction errors of all normal samples to form a normal error distribution; calculate its quantiles: if the error exceeds 99% of the normal samples, the anomaly probability is 0.99; if it exceeds 99.9% of the normal samples, the probability is 0.999.

[0104] S311. Obtain the historical baseline risk coefficient of the road level corresponding to the route segment that matches each time window in the current transportation task of the financial customer; multiply the abnormal risk probability value corresponding to each time window in the current transportation task of the financial customer by the historical baseline risk coefficient of the corresponding road level to obtain the weighted abnormal score of each time window.

[0105] Specifically, the maximum and minimum historical accident rates are found across all road grades. For each road grade, the difference between the historical accident rate and the minimum historical accident rate, and the difference between the maximum and the minimum historical accident rates are divided to obtain the historical baseline risk coefficient for the corresponding road grade.

[0106] S312. Obtain the maximum and minimum values ​​of the weighted anomaly scores for all time windows; use the difference between the maximum and minimum values ​​as the weighted anomaly score range; divide the difference between the weighted anomaly score and the minimum value for each time window by the weighted anomaly score range to obtain the transportation process anomaly risk for each time window; perform an arithmetic average of the cargo status anomaly risks for all time windows in the current transportation task of the financial customer to obtain the cargo status anomaly risk for the current transportation task of the financial customer.

[0107] In this embodiment, the construction process of the preset risk fusion assessment function in step S4 includes the following specific steps:

[0108] S41. Extract the potential energy for cargo compensation triggering the current transportation of goods by financial clients and the abnormal cargo status risk of the current transportation task;

[0109] S42. Construct a two-player zero-sum game model, taking the cargo compensation triggering potential and the cargo state abnormality risk of the current transportation task as the payoffs of the two players; solve the mixed strategy Nash equilibrium through an iterative algorithm to obtain the optimal weight w of the cargo compensation triggering potential in the fusion evaluation and the optimal weight 1-w of the cargo state abnormality risk of the current transportation task.

[0110] The specific steps are as follows: Extraction of the compensation trigger potential factor: This step aims to generate a primary quantitative indicator characterizing the severity of the potential financial risks of a transportation mission. Specifically, it involves the following operations:

[0111] Data Acquisition: Extract static and historical data related to the current transportation task from the associated order management subsystem, insurance business subsystem, and historical claims database. The data should include at least: declared value of goods, insurance policy information (including insured amount and deductible clauses), shipper identification, and goods classification code; Feature Analysis: Based on the data, analyze the following features: inherent value level of goods, adequacy of insurance coverage, contractually agreed risk-bearing threshold (i.e., deductible), and historical claim frequency and intensity of the shipper or similar goods; Factor Calculation: Through a preset weighted or aggregation algorithm, integrate the above feature analysis results to output a scalar value, called the cargo claim triggering potential. This value is positively correlated with cargo value, insurance coverage, and historical claim tendency, and negatively correlated with the deductible. This factor essentially quantifies the probability and expected loss amount of financial compensation from the insurance company once an insured event occurs; Abnormal Status Risk Factor Extraction: This step aims to generate a second quantitative indicator characterizing the immediate probability of physical damage during transportation; it specifically performs the following operations:

[0112] Data Acquisition: Extract dynamic and attribute data related to the current transportation process in real-time or near real-time from the Internet of Things (IoT) sensing subsystem, Geographic Information System (GIS), and carrier management database. The data should include at least: vehicle sensor data (temperature, humidity, acceleration, tilt angle), real-time location and planned route of the vehicle, real-time traffic and weather data, historical performance data of the carrier (or driver) (e.g., on-time delivery rate, damage rate), and cargo physical sensitivity classification.

[0113] Feature analysis: Based on the data, analyze the following features: the deviation between the transportation environment parameters and the suitable range of the goods, the route risk level (based on weather, road conditions, and accident hotspots), the carrier reliability score, and the vulnerability level of the goods themselves.

[0114] Factor calculation: Through another pre-defined weighted or aggregated algorithm, the results of the above dynamic and attribute feature analysis are integrated and calculated to output a scalar value called cargo condition anomaly risk. This value is positively correlated with environmental deviation, route risk, carrier unreliability, and cargo vulnerability. This factor essentially quantifies the real-time probability of cargo suffering physical damage or condition anomalies during transportation.

[0115] Technical effect: Through the above method, the system extracts two orthogonal and complementary risk quantification factors in parallel for the same transportation task: one from the perspective of the severity of financial consequences, and the other from the perspective of the probability of damage occurring in transit; this provides an accurate and multi-dimensional data foundation for subsequent comprehensive and balanced risk assessment.

[0116] S42: The specific steps of the adaptive risk factor fusion weight determination method based on game equilibrium are as follows: Game model construction: Establish a two-person zero-sum game model; where:

[0117] Participant 1 is defined as a financial risk perspective, and its strategic influence comes from the cargo compensation triggering potential factor; Participant 2 is defined as a physical risk perspective, and its strategic influence comes from the cargo abnormality risk factor.

[0118] Game objective: The goal of the competition between the two sides is to maximize the weight ratio of their respective factors in the final comprehensive risk score under the constraint that the total weight sum is a fixed value (normalized to 1); an increase in the weight of one side will inevitably lead to a decrease in the weight of the other side, forming a zero-sum relationship.

[0119] Iterative weighting (Nash equilibrium search): An iterative algorithm is used to simulate the game process to find the Nash equilibrium point, and the weight allocation corresponding to this point is the optimal solution; the iterative process is as follows:

[0120] Initialization: Assign initial weights to the two factors (e.g., 0.5 each);

[0121] Benefit Evaluation: In each iteration, calculate the relative benefit gain for a particular factor if the weights are fine-tuned to favor it more under the current weight allocation; this gain is positively correlated with the significance of the current value of that factor relative to another factor; specifically:

[0122] If the value of the cargo compensation trigger potential factor is significantly higher than that of the cargo abnormality risk factor, it indicates that the severity of financial risk is exceptionally prominent. In this case, increasing the weight of the financial risk perspective will yield a significant benefit.

[0123] Conversely, if the value of the abnormal cargo condition risk factor is significantly higher, it indicates that the urgency of physical damage is exceptionally prominent. In this case, increasing the weight of the physical risk perspective will yield a significant benefit.

[0124] Strategy Adjustment (Weight Update): Based on the relative numerical ratio of the factors on both sides, the weight allocation is adjusted according to the preset update rules (such as gradient or response function); the adjustment direction is: to give the factor with the current relatively higher value a greater tendency to increase its weight;

[0125] Convergence judgment: Repeat the above evaluation and adjustment process until a stable state is reached; in this state, neither party can gain additional benefits by unilaterally changing the weight allocation (i.e., changing the strategy); this stable state is the Nash equilibrium of the game; Optimal weight output: The weight allocation scheme obtained after iterative convergence is used as the final fusion weight of the two risk factors corresponding to the current transportation task.

[0126] S43. Based on the formula: compensation risk = w × cargo compensation trigger potential energy + (1-w) × cargo status abnormality risk of the current transportation task, calculate the compensation risk and cargo compensation trigger potential energy of the current transportation task for the financial customer.

[0127] In this embodiment, step S5, based on the risk assessment results of the financial customer's current transportation task, provides a risk warning for the financial customer's goods as the insured party, specifically including:

[0128] S51. Obtain the risk assessment results of the compensation for the current transportation task of the financial customer;

[0129] S52. A preset compensation risk threshold is established. When the compensation risk assessment result of a financial customer's current transportation task exceeds the compensation risk threshold, a cargo compensation warning is issued for the current transportation task, and emergency maintenance is performed on the cargo associated with the current transportation task. When the compensation risk assessment result of a financial customer's current transportation task is less than or equal to the compensation risk threshold, the current transportation task is marked as completed. It should be noted that the weights and thresholds in this embodiment are determined as follows: the main attribute data and historical financial transaction data of the corresponding financial customers in 3000 sets of transportation tasks, as well as the cargo transportation data and transportation route data of the cargo associated with the corresponding financial customers, are obtained to calculate the compensation risk of the transportation task. At the same time, the judgment results of whether cargo damage and compensation have occurred in the 3000 sets of transportation tasks are obtained. The compensation risks of the 3000 sets of transportation tasks and the corresponding judgment results of whether cargo damage and compensation have occurred are imported into the fitting software for fitting, and the corresponding weights and thresholds that meet the highest determination coefficient are output.

[0130] Example 2

[0131] like Figure 3 As shown, this embodiment provides a financial customer data management system based on big data assessment, including:

[0132] The data acquisition module is used to acquire the main attribute data and historical financial transaction data of the financial customer who is the insured party, and at the same time acquire the cargo transportation data and transportation route data of the goods associated with the financial customer who is the insured party.

[0133] The transportation claims analysis module is used to extract the task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and a binary indicator of whether a claim has been made for the corresponding transportation task from historical transportation task records completed in historical financial transaction data. Simultaneously, it extracts the historical claim rate for the corresponding customer from the main attribute data, constructs a historical damage claims dataset, and statistically calculates the percentage of records with claims made under each pairwise combination of cargo type code and damage location code, defining this as the baseline functional influence weight for each pairwise combination of cargo type code and damage location code. Therefore, through the historical damage claims dataset and the baseline functional influence weight, a cargo claim triggering potential prediction model is constructed to analyze the cargo claim triggering potential of financial customers.

[0134] The cargo status detection module divides the total duration of each transportation task in the historical normal cargo status dataset into multiple time windows. For each time window, it extracts position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical attenuation risk coefficients representing the mechanical performance degradation of the transport vehicle from the cargo transportation data. These parameters characterize the transportation status of each time window. Using these transportation status parameters from the historical normal cargo status dataset, a cargo status anomaly risk detection model based on the nonlinear coupling relationship of multidimensional time series transportation status parameters is trained. This model is used to determine the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window based on a comprehensive analysis of the five transportation status parameters. Based on the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window output by the cargo status anomaly risk detection model and the historical benchmark risk coefficients matching each time window in the transportation route data, the module analyzes the cargo status anomaly risk of the current transportation task for the financial client.

[0135] The claims risk detection module is used to input the analysis results of the cargo claims triggering potential of financial customers and the analysis results of the abnormal cargo status risk of the current transportation task of financial customers into a preset risk fusion assessment function based on dynamic game theory, and to assess the claims risk of the current transportation task of financial customers and the cargo claims triggering potential.

[0136] The claims warning module is used to issue claims warnings for the goods of financial clients who are the insured parties, based on the claims risk assessment results of the current transportation task.

[0137] The steps for implementing the corresponding functions of each parameter and unit module in the financial customer data management system based on big data assessment of the present invention described above can be referred to the parameters and steps in the embodiments of the financial customer data management method based on big data assessment described above, and will not be repeated here.

[0138] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the embodiments for IoT devices and media are relatively simple in description because they are fundamentally similar to the method embodiments; relevant parts can be referred to the descriptions in the method embodiments.

[0139] The systems, media, and methods provided in the embodiments of the present invention are in one-to-one correspondence. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0140] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0141] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0142] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0143] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0144] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0145] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0146] It should also be noted that the terms include, encompass, or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitations, the inclusion of an element by a statement that includes a… does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0147] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A financial customer data management method based on big data assessment, characterized in that, Includes the following steps: S1. Obtain the main attribute data and historical financial transaction data of the financial customer who is the insured party, and at the same time obtain the cargo transportation data and transportation route data of the goods associated with the financial customer who is the insured party. S2. Based on the historical transportation task records completed in historical financial transaction data, extract the task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and binary indicators of whether a claim was made for the corresponding transportation task from the historical transportation task records; at the same time, extract the historical claim rate of the corresponding customer from the main attribute data, construct a historical damage claim dataset, and based on the historical damage claim dataset, statistically calculate the percentage of records with claims made under each pairwise combination of cargo type code and damage location code to the total number of records, define it as the baseline function influence weight corresponding to each pairwise combination of cargo type code and damage location code; thereby, through the historical damage claim dataset and the baseline function influence weight, construct a cargo claim trigger potential prediction model to analyze the cargo claim trigger potential of financial customers; S3. Divide the total duration of each transportation task in the historical normal cargo status dataset into multiple time windows. For each time window, extract the position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical attenuation risk coefficient (characterizing the mechanical performance degradation of the transport vehicle) from the cargo transportation data as transportation status parameters for each time window. Using the transportation status parameters of each time window in the historical normal cargo status dataset, train a cargo status anomaly risk detection model based on the nonlinear coupling relationship of multi-dimensional time series transportation status parameters. This model is used to determine the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window based on the comprehensive analysis of the five transportation status parameters. Based on the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window output by the cargo status anomaly risk detection model and the historical benchmark risk coefficient matching each time window in the transportation route data, analyze the cargo status anomaly risk of the current transportation task for the financial customer. Among them, the position change parameter is used to identify abnormal position changes and detect the risk of path deviation; the speed fluctuation parameter and driving aggressive parameter are used to identify driving behavior; the environmental deviation parameter is used to determine whether the cargo storage conditions are compliant; the sealing status parameter is used to detect whether the cargo has been illegally touched; and the mechanical deterioration status risk coefficient, which characterizes the mechanical performance degradation of the transport vehicle itself, is used to identify the potential transportation interruption or cargo loss risk caused by the deterioration of the technical condition of the transport vehicle. S4. Input the analysis results of the potential energy for cargo compensation triggering the financial customer and the analysis results of the abnormal cargo status risk of the current transportation task of the financial customer into the preset risk fusion assessment function based on dynamic game theory to assess the compensation risk of the current transportation task of the financial customer. S5. Based on the risk assessment results of the current transportation task of the financial customer, issue a claim warning for the goods of the financial customer who is the insured party.

2. The financial customer data management method based on big data assessment according to claim 1, characterized in that, In step S2, a cargo compensation trigger potential prediction model is constructed by using historical damage claims datasets and benchmark function influence weights to analyze the cargo compensation trigger potential of financial customers. The specific steps include the following: S23. Based on the logistic regression model and the historical damage claims dataset, construct and train a cargo compensation trigger potential prediction model; S24. For the goods currently being transported by the financial customer, obtain the damage location code, damage area percentage, and average depth change value of the damaged parts of the goods through image recognition and 3D scanning, and obtain the goods type code of the goods; based on the goods type code and damage location code, query the baseline function influence weight corresponding to the pairwise combination of the goods type code and damage location code, and obtain the historical claim rate of the corresponding customer; input the damage area percentage, average depth change value of the damaged parts, the corresponding baseline function influence weight, the historical claim rate of the corresponding customer, and the goods type code of the goods currently being transported into the trained goods claim trigger potential prediction model; the model outputs the trigger claim probability value of the goods currently being transported, and uses the trigger claim probability value of the goods currently being transported as the goods claim trigger potential value of the financial customer's goods currently being transported.

3. The financial customer data management method based on big data assessment according to claim 2, characterized in that, Step S3 involves extracting position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical degradation risk coefficient (characterizing the degradation of the transport vehicle's mechanical performance) from the cargo transport data for each time window. This process includes the following specific steps: S31. From the cargo transportation data, obtain the vehicle latitude and longitude coordinate sequence and instantaneous speed sequence of the transport vehicle corresponding to the transportation task; the longitudinal and lateral acceleration event signal sequence exceeding the preset intensity threshold collected through the vehicle controller local area network bus; the temperature reading sequence and relative humidity reading sequence collected through the environmental sensors fixed inside the cargo compartment; the electronic seal status signal sequence collected through the electronic seal installed at the cargo unit door; and the raw diagnostic data stream reflecting the working status of each subsystem of the vehicle collected in real time through the vehicle self-diagnosis system interface, wherein the electronic seal status signal includes a locked status signal and an unlocked status signal; from the transportation route data, obtain the route segment sequence associated with the digital map road network, and the historical benchmark risk coefficient of the road level corresponding to each route segment, and dynamically update it after feature fusion based on the historical accident rate of the corresponding road segment; S32. Divide the transportation task into multiple time windows with equal time intervals according to the duration of the transportation task, and obtain the cargo transportation data and transportation route data corresponding to each time window in the transportation task. S33. Calculate the spherical linear displacement distance based on the vehicle's latitude and longitude coordinates at the beginning and end of the corresponding time window, and use it as the position change parameter for the corresponding time window; S34. Calculate the standard deviation of all instantaneous velocity reading sequences within the corresponding time window, and use it as the velocity fluctuation parameter for the corresponding time window; S35. The total number of longitudinal and lateral acceleration event signals triggered within the corresponding time window is accumulated and used as the driving aggressive parameter for the corresponding time window. S36. Determine whether the reading in any of the temperature reading sequence and relative humidity reading sequence within the corresponding time window exceeds the preset safety threshold range. If it does, set the environmental deviation parameter of the corresponding time window to 1; otherwise, set the environmental deviation parameter of the corresponding time window to 0. S37. Identify whether the electronic seal status signal sequence within the corresponding time window changes from a locked status signal to an unlocked status signal. If it does, assign the sealing status parameter of the corresponding time window to 1; otherwise, assign it to 0. S38. Extract features from the raw diagnostic data streams that reflect the working status of each subsystem of the vehicle collected within the corresponding time window, and calculate the degree of mechanical performance degradation of the transport vehicle within the current time window based on the extracted feature values, which serves as the mechanical degradation state risk coefficient characterizing the mechanical performance degradation of the transport vehicle itself.

4. The financial customer data management method based on big data assessment according to claim 3, characterized in that, Step S3 involves analyzing the risk of abnormal cargo status for the current transportation task of the financial customer, including the following specific steps: S39. Construct a transportation state parameter coupling analysis model based on deep neural network. Take the transportation state parameter vector of each time window in the historical cargo state normal dataset as input and train it with the goal of reconstructing the corresponding vector. This enables the model to learn the nonlinear coupling relationship between various transportation state parameters under normal conditions. S310. Input the transportation status parameter vectors of all time windows in the current transportation task of the financial customer into the trained transportation status parameter coupling analysis model, calculate the reconstruction error of the model to the input vector, use the reconstruction error as a measure of the deviation of the nonlinear coupling relationship between each transportation status parameter in the corresponding time window from the normal state, and convert the corresponding reconstruction error into an abnormal risk probability value as the probability of an abnormal risk of unauthorized path deviation occurring in the corresponding time window. S311. Obtain the historical baseline risk coefficient of the road level corresponding to the route segment that matches each time window in the current transportation task of the financial customer; multiply the abnormal risk probability value corresponding to each time window in the current transportation task of the financial customer by the historical baseline risk coefficient of the corresponding road level to obtain the weighted abnormal score of each time window. S312. Obtain the maximum and minimum values ​​of the weighted anomaly scores for all time windows; use the difference between the maximum and minimum values ​​as the weighted anomaly score range; divide the difference between the weighted anomaly score and the minimum value for each time window by the weighted anomaly score range to obtain the transportation process anomaly risk for each time window; perform an arithmetic average of the cargo status anomaly risks for all time windows in the current transportation task of the financial customer to obtain the cargo status anomaly risk for the current transportation task of the financial customer.

5. The financial customer data management method based on big data assessment according to claim 4, characterized in that, The construction process of the preset risk fusion assessment function in step S4 includes the following specific steps: S41. Extract the potential energy for cargo compensation triggering the current transportation of goods by financial clients and the abnormal cargo status risk of the current transportation task; S42. Construct a two-player zero-sum game model, taking the cargo compensation triggering potential and the cargo state abnormality risk of the current transportation task as the payoffs of the two players; solve the mixed strategy Nash equilibrium through an iterative algorithm to obtain the optimal weight w of the cargo compensation triggering potential in the fusion evaluation and the optimal weight 1-w of the cargo state abnormality risk of the current transportation task. S43. According to the compensation risk = w × cargo compensation trigger potential energy + (1-w) × cargo status abnormality risk of the current transportation task; The compensation risk and the potential energy triggered by the compensation of goods for the current transportation task of the financial customer are calculated.

6. The financial customer data management method based on big data assessment according to claim 5, characterized in that, In step S5, based on the risk assessment results of the financial customer's current transportation task, a claim warning is issued for the goods of the financial customer, who is the insured party. Specifically, this includes: S51. Obtain the risk assessment results of the compensation for the current transportation task of the financial customer; S52. Preset compensation risk threshold: When the compensation risk assessment result of the financial customer's current transportation task is greater than the compensation risk threshold, issue a cargo compensation warning for the financial customer's current transportation task and perform emergency maintenance on the cargo associated with the current transportation task; when the compensation risk assessment result of the financial customer's current transportation task is less than or equal to the compensation risk threshold, mark the current transportation task as a completed transportation task.

7. A financial customer data management system based on big data assessment, implemented according to any one of claims 1-6, characterized in that, The system includes: The data acquisition module is used to acquire the main attribute data and historical financial transaction data of the financial customer who is the insured party, and at the same time acquire the cargo transportation data and transportation route data of the goods associated with the financial customer who is the insured party. The transportation claims analysis module is used to extract the task number, cargo type code, damage location code, damage area percentage, average depth change of the damaged area, and a binary indicator of whether a claim has been made for the corresponding transportation task from historical transportation task records completed in historical financial transaction data. Simultaneously, it extracts the historical claim rate for the corresponding customer from the main attribute data, constructs a historical damage claims dataset, and statistically calculates the percentage of records with claims made under each pairwise combination of cargo type code and damage location code, defining this as the baseline functional influence weight for each pairwise combination of cargo type code and damage location code. Therefore, through the historical damage claims dataset and the baseline functional influence weight, a cargo claim triggering potential prediction model is constructed to analyze the cargo claim triggering potential of financial customers. The cargo status detection module divides the total duration of each transportation task in the historical normal cargo status dataset into multiple time windows. For each time window, it extracts position change parameters, speed fluctuation parameters, driving aggression parameters, environmental deviation parameters, sealing status parameters, and mechanical attenuation risk coefficients representing the mechanical performance degradation of the transport vehicle from the cargo transportation data. These parameters characterize the transportation status of each time window. Using these transportation status parameters from the historical normal cargo status dataset, a cargo status anomaly risk detection model based on the nonlinear coupling relationship of multidimensional time series transportation status parameters is trained. This model is used to determine the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window based on a comprehensive analysis of the five transportation status parameters. Based on the probability of unauthorized path deviation anomaly risk occurring within the corresponding time window output by the cargo status anomaly risk detection model and the historical benchmark risk coefficients matching each time window in the transportation route data, the module analyzes the cargo status anomaly risk of the current transportation task for the financial client. The claims risk detection module is used to input the analysis results of the cargo claims triggering potential of financial customers and the analysis results of the abnormal cargo status risk of the current transportation task of financial customers into a preset risk fusion assessment function based on dynamic game theory, and to assess the claims risk of the current transportation task of financial customers and the cargo claims triggering potential. The claims warning module is used to issue claims warnings for the goods of financial clients who are the insured parties, based on the claims risk assessment results of the current transportation task.

Images