Multi-factor integration and optimization method for dynamic risk pricing of insurance products

Abstract

The present application belongs to the technical field of insurance product pricing, and more specifically, relates to a method for optimizing dynamic risk pricing of insurance products by integrating multiple factors; collecting multi-source information data and performing data preprocessing; constructing a spatiotemporal cumulative risk index based on travel trajectory analysis of on-board OBD data, combining high-risk road sections, driving behavior, and environmental factors; using a graph convolutional network to model the risk transfer effect in the social circle of car owners, capturing the mutual influence between car owners in the social circle; using a neural network based on multi-task learning to comprehensively process dynamic risks, social risk characteristics, and other relevant data; optimizing the prediction of the probability of accidents and the amount of losses through a shared network layer; and generating insurance pricing in combination with external market data to ensure the comprehensiveness and accuracy of insurance pricing; solving the blind spots in traditional pricing methods, making insurance pricing more intelligent and accurate, and able to reflect the actual risk status of car owners in real time.

Description

Technical Field

[0001] The present application belongs to the technical field of insurance product pricing, and more specifically, relates to a method for optimizing dynamic risk pricing of insurance products by integrating multiple factors. Background Art

[0002] The insurance industry faces the challenge of accurately assessing risk and establishing reasonable pricing during the pricing process. Traditional insurance pricing methods are primarily based on historical data and statistical models. However, with the surge in data volumes and the continuous advancement of technology, insurance product pricing urgently requires more intelligent and precise solutions.

[0003] Traditional insurance pricing methods typically rely on static characteristics, such as basic information like the owner's age, gender, and driving history. However, these static characteristics fail to fully account for the owner's dynamic risk factors, such as real-time driving behavior, vehicle location, and road conditions. This makes traditional pricing models inadequate for reflecting the owner's actual risk profile. With the advancement of in-vehicle technology, the widespread use of on-board diagnostics (OBD) devices and real-time weather data can provide real-time and detailed driving data and environmental information, adding a new dimension to insurance pricing.

[0004] Furthermore, existing risk assessment methods often overlook the social connections and risk transmission effects between car owners. Car owners within a social circle may influence each other to a certain extent. For example, the risk of certain driving behaviors can have a ripple effect within the social circle, affecting individuals throughout the entire social network. This factor is a significant blind spot for traditional pricing models.

[0005] However, the current conventional pricing method is to price based on the owner's insurance claims and vehicle volume. This not only ignores the dynamic risks of the owner himself, but also leads to a one-sided pricing method, which in turn causes the final insurance pricing to fail to match the actual situation. Summary of the Invention

[0006] The present invention provides a dynamic risk pricing optimization method for insurance products that integrates multiple factors, aiming to solve the technical problems that the current pricing method is one-sided and does not consider the dynamic risks of the car owner himself, resulting in the final insurance pricing not matching the actual situation.

[0007] The multi-factor integrated dynamic risk pricing optimization method for insurance products includes the following steps:

[0008] Step 1: Collect the car owner's insurance policy records, claims data, vehicle OBD data, meteorological data, and geographic information data, and pre-process the collected data;

[0009] Step 2: Obtain the driver's travel trajectory based on the vehicle's OBD data. Based on the travel trajectory and high-risk road sections, the driver's overall road section risk is constructed. The driver's driving behavior and environmental factors are then incorporated to obtain the spatiotemporal cumulative risk index.

[0010] Step 3: Build a graph structure with the car owner as the node and the car owner's social circle as the edge between the nodes. Use the graph convolutional network to perform convolution operations on the graph structure to learn the risk transmission effect within the social circle and generate social risk characteristics;

[0011] Step 4: Build a multi-task learning neural network based on the preprocessed data, social risk characteristics, and cumulative risk index. Use the shared network layer to learn the common features of the probability of accident and loss amount, and output the probability of accident and the predicted loss amount.

[0012] Step 5: Generate insurance pricing based on the risk score and predicted loss amount output by the multi-task learning neural network and external market data.

[0013] The present invention collects vehicle owner insurance policy records, claims data, on-board optical disc (OBD) data, meteorological data, and geographic information data for data preprocessing, providing comprehensive and high-quality data support for subsequent risk assessment. Secondly, through travel trajectory analysis based on on-board OBD data, combined with high-risk road sections, driving behavior, and environmental factors, a spatiotemporal cumulative risk index is constructed. This index considers the impact of vehicle owners' dynamic driving behavior and the real-time environment, addressing the problem that traditional pricing models fail to fully account for vehicle owners' dynamic risks. Furthermore, a graph convolutional network is used to model the risk transmission effect within vehicle owners' social circles, effectively capturing the mutual influence between vehicle owners within the social circle, addressing the problem that traditional models fail to account for social risk transmission. Finally, a multi-task learning-based neural network comprehensively processes dynamic risk, social risk characteristics, and other relevant data, optimizes predictions of accident probability and loss amount through a shared network layer, and generates insurance pricing based on external market data, ensuring comprehensive and accurate insurance pricing. Therefore, through multi-factor fusion, dynamic risk assessment, and social circle modeling, this technical solution addresses the blind spots existing in traditional pricing methods, making insurance pricing more intelligent and accurate, and able to reflect the actual risk status of vehicle owners in real time.

[0014] Preferably, the step 2 comprises the following steps:

[0015] High-risk road section calibration: Determine the accident risk value of each road section through historical data and real-time traffic monitoring:

[0016]

[0017] Where: λ represents the accident attenuation coefficient; N acc(i, t) represents the number of accidents on road section i at time t; τ represents the current timestamp; k represents the time window size; R acc (i, t) represents the accident risk value of road section i at time t; e represents the base of the natural logarithm;

[0018] Dynamic risk path allocation: For each trip, the risk weight of the driver on different road sections is calculated:

[0019]

[0020] Where: Δt i Indicates the time the driver travels on this road section; R path (P) represents the overall path risk of the driver’s driving path;

[0021] Driving behavior correction factor calculation: Risky driving behavior is corrected by multiplying the current risk assessment value with a weighting factor:

[0022]

[0023] Where: α brake Indicates the weight factor of emergency braking; S brake (i, t) represents the emergency braking intensity at time t; S max Indicates the maximum value of emergency braking;

[0024] Weather Modifier:

[0025] f weather (i, t) = β1·f precip (i, t)+β2·f fog (i, t);

[0026] Where: represents the precipitation impact factor; represents the haze impact factor; P(i, t) represents the rainfall in region i at time t; V(i, t) represents the visibility in region i at time t; β1 and β2 represent weighting factors;

[0027] Road condition correction factor:

[0028]

[0029] Where: Q(i, t) represents the traffic flow of area i at time t; Q max (i) represents the historical maximum flow in area i;

[0030] Calculation of spatiotemporal cumulative risk index:

[0031] R driver (i, t) = R acc (i, t)·(1+ΔRbrake (i, t)+f weather (i, t)+f trafffc (i, t));

[0032] Where: ΔR brake (i, t) represents the correction factor for sudden braking; f weather (i, t) represents the weather correction factor; f traffic (i, t) represents the traffic flow correction factor.

[0033] Preferably, the graph structure is set as G = (V, E, X), where V represents a node set, each node represents a car owner; E represents an edge set, edges represent social relationships between car owners, and X represents a node feature matrix, each node contains the risk characteristics of the car owner; wherein the node feature matrix X contains the initial risk feature vector of each car owner; the initial risk feature vector includes the overall path risk R of car owner i path (P), emergency brake correction factor ΔR brake (i, t), weather correction factor f weather (i, t) and traffic flow correction factor f traffic (i, t); where the adjacency matrix A represents the social relationship between car owners, where the value is 1 if there is a social relationship and 0 if there is no social relationship.

[0034] Preferably, the graph convolutional network includes a first data processing layer, a second data processing layer and a global pooling layer;

[0035] The first data processing layer includes a graph convolution layer and a multi-layer perceptron nonlinear transformation layer. The graph convolution layer performs a convolution operation on the input feature matrix to convert it into a first feature matrix. The first feature matrix is ​​used as the input of the multi-layer perceptron nonlinear transformation layer. The multi-layer perceptron nonlinear transformation layer performs a nonlinear transformation on the first feature matrix to obtain the feature matrix output by the first layer.

[0036] The second data processing layer includes a graph convolution layer and an attention mechanism layer. The graph convolution layer performs a convolution operation on the feature matrix output by the first layer to obtain a second feature matrix. The attention mechanism layer calculates attention weights based on the output second feature matrix, and obtains a feature matrix of the second layer output weighted by the attention mechanism layer based on the calculated attention weights and the second feature matrix.

[0037] Global pooling layer: Aggregates the features of all nodes in the graph structure to generate a global representation of the graph and obtain social risk features.

[0038] Preferably, the loss function of the graph convolutional network is as follows:

[0039]

[0040] Where: regularization and λ separation Both represent weight parameters;

[0041]

[0042] Where: N represents the number of graphs; R i represents the true social risk characteristics of the i-th graph; represents the predicted social risk characteristics of the i-th graph; represents the mean square error;

[0043]

[0044] Where: represents the global representation of the i-th graph; represents the regularization loss;

[0045]

[0046] Where: (R i -R j ) 2 Used to weight the distance between global representations; represents the global representation of the jth graph; R j represents the true social risk feature of the j-th graph; ||·||2 represents the Euclidean distance.

[0047] Preferably, the multi-task neural network includes:

[0048] Input layer: The input vector includes social risk characteristics, spatiotemporal cumulative risk index, and preprocessed data;

[0049] Shared network layer: A multi-layer perceptron is used as the shared network layer, and the input vector is processed through the shared network layer to obtain shared features;

[0050] Accident probability prediction layer: The shared features are processed through the accident probability layer to obtain the probability of occurrence:

[0051]

[0052] Where: W prob b represents the weight matrix for accident probability prediction; prob F represents the bias term of the accident probability prediction layer; shared represents shared features; σ represents the sigmoid activation function; Indicates the predicted probability of danger;

[0053] Loss amount prediction layer: The loss amount prediction layer processes shared features to obtain the loss amount prediction value:

[0054]

[0055] Where: Indicates the predicted loss amount; W loss b represents the weight matrix of loss amount prediction; loss Represents the bias term of the loss amount prediction layer.

[0056] Preferably, the loss function of the multi-task neural network is as follows:

[0057]

[0058] Where: risk The weight parameter representing the probability of loss; To express the loss of the probability of accident, the binary cross entropy loss function is used; Indicates the loss of the loss amount, using mean square error loss; λ loss The weight parameter representing the loss amount.

[0059] Preferably, the step 5 comprises the following steps:

[0060] Basic pricing: The basic pricing is determined based on the predicted probability of loss and the predicted loss amount:

[0061]

[0062] Where: Indicates the predicted probability of danger; Indicates the predicted loss amount; P base Indicates base pricing;

[0063] Base pricing adjustment: Adjust the base pricing based on market factors to obtain the adjusted insurance pricing:

[0064] P final =P base ×(1+α×ΔR+β×C market +γ×I inflation );

[0065] Where: ΔR represents the difference from the industry benchmark rate; C market Indicates market competition status; I inflation represents the current inflation rate; α, β, and γ represent adjustment coefficients.

[0066] Preferably, in step 5, the adjustment coefficients α, β and γ are adjusted by defining a loss function:

[0067]

[0068] Where: represents the insurance pricing after the i-th adjustment; represents the actual insurance pricing in history; the optimal adjustment coefficient is found by minimizing the loss function.

[0069] The beneficial effects of the present invention include:

[0070] By collecting a variety of data, the present invention comprehensively reflects the car owner's historical behavior, real-time driving conditions and external environmental factors, building a solid foundation for subsequent risk assessment and solving the problem of single and one-sided data sources in traditional pricing models.

[0071] The present invention obtains the driver's travel trajectory through on-board OBD data, and combines high-risk sections, driving behavior and environmental factors to construct a spatiotemporal cumulative risk index. This fully considers the driver's dynamic driving behavior and real-time environmental impact, and solves the problem of traditional pricing models ignoring the driver's dynamic risk. By combining high-risk sections with specific travel trajectories, the risk status of the driver in different sections and at different times can be more accurately assessed.

[0072] This paper takes car owners as nodes and introduces their social circles as edges between nodes. It uses graph convolutional networks to perform convolution operations, captures the risk transmission effect within the car owner's social circle, and generates social risk features. Based on this, it can identify the mutual influence between car owners, especially the transmission effect of risky behaviors within the social circle, solving the problem of traditional models ignoring social risk transmission, thereby more comprehensively evaluating the actual risks of car owners.

[0073] The present invention uses a multi-task learning neural network to learn the common characteristics of the probability of accident and the amount of loss through a shared layer, and outputs the probability of occurrence and the predicted amount of loss respectively. This not only improves the learning efficiency of the model, but also can better capture the potential correlation between the two tasks, solves the suboptimal problem caused by separation, and improves the accuracy and robustness of risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0075] Figure 1 This is a flowchart of the overall steps provided by an embodiment of the present invention.

[0076] Figure 2 A schematic diagram illustrating the structure of a graph convolutional network provided by an embodiment of the present invention.

[0077] Figure 3 A schematic diagram illustrating the structure of a neural network for multi-task learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0078] In order to make the technical problems, technical solutions and beneficial effects to be solved by this application more clearly understood, this application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0079] See also Figure 1 As shown in FIG, the multi-factor integrated dynamic risk pricing optimization method for insurance products includes the following steps:

[0080] Step 1: Collect the car owner's insurance policy records, claims data, vehicle OBD data, meteorological data, and geographic information data, and pre-process the collected data;

[0081] Exemplary:

[0082] Car owner's policy record: Obtain the car owner's policy information through the insurance company system, including policy number, insurance type, historical policy renewal status, coverage, insurance amount, policy effective and expiration dates, etc.

[0083] Claims data: Obtain the car owner's historical claims records from the insurance company, including the time of each accident, accident type, claim amount, responsible party, geographical location of the accident, and road conditions.

[0084] On-board OBD data: Real-time driving data is obtained through the OBD (on-board diagnostic) device of the owner's vehicle, including vehicle speed, acceleration, braking conditions, engine speed, fuel consumption, mileage, driving time, driving habits (such as sudden braking, sudden acceleration, turning, etc.), fault codes, etc.

[0085] Meteorological data: Real-time weather information related to the owner's travel is collected through meteorological data interfaces (such as weather service APIs), including temperature, humidity, precipitation, wind speed, visibility, etc.

[0086] Geographic information data: Use the Geographic Information System (GIS) to obtain information such as the driver's travel route, the geographical coordinates of the road, high-risk sections (such as sections with frequent traffic accidents, landslide areas, etc.), road grade, traffic flow, etc.

[0087] Data cleaning:

[0088] Filter invalid data: such as filling in missing values ​​and deleting invalid or incomplete records.

[0089] Outlier detection: Identify and correct abnormal data by setting reasonable ranges (e.g., vehicle speed should not exceed a certain range).

[0090] Data deduplication: Ensure that only one piece of data for the same time period and the same event is retained, and remove duplicate data.

[0091] Data standardization and normalization:

[0092] All numerical data (such as vehicle speed, mileage, weather conditions, etc.) are standardized so that data of different dimensions fluctuate within the same range, avoiding the impact of scale differences on model performance due to certain features.

[0093] Use techniques such as Z-score normalization (mean is 0, standard deviation is 1) or Min-Max normalization (data is mapped to the interval [0, 1]) to unify the distribution of data.

[0094] Time Series Processing:

[0095] Since vehicle OBD data and weather data are usually collected in time series, it is necessary to organize the data in chronological order according to the driver's specific travel trajectory and process the driver's driving behavior and weather conditions in different time periods.

[0096] Handling missing time series data: Fill in missing time period data through interpolation to ensure data integrity.

[0097] Feature extraction:

[0098] We extract driving behavior features from the vehicle's OBD data, including average speed, number of sudden brakes, driving distance, and driving time.

[0099] Extract weather impact characteristics from meteorological data, such as the impact of rainfall and wind speed on the probability of accidents;

[0100] Features related to road risks are extracted from geographic information data, such as the frequency of vehicle owners driving on high-risk sections, section types (urban roads, highways, etc.), road condition information, etc.

[0101] Integrate car owners' insurance policy records, claims data, OBD data, meteorological data and geographic information data to form a comprehensive data set containing all relevant information.

[0102] Convert data from different sources into a unified format (for example, unified timestamp format, standardization of coordinate systems) to ensure data consistency and availability.

[0103] Step 2: Obtain the driver's travel trajectory based on the vehicle's OBD data. Based on the travel trajectory and high-risk road sections, the driver's overall road section risk is constructed. The driver's driving behavior and environmental factors are then incorporated to obtain the spatiotemporal cumulative risk index.

[0104] The step 2 comprises the following steps:

[0105] High-risk road section calibration: Determine the accident risk value of each road section through historical data and real-time traffic monitoring:

[0106]

[0107] Where: λ represents the accident attenuation coefficient; N acc (i, t) represents the number of accidents on road section i at time t; τ represents the current timestamp; k represents the time window size; R acc (i, t) represents the accident risk value of road section i at time t; e represents the base of the natural logarithm;

[0108] Dynamic risk path allocation: For each trip, the risk weight of the driver on different road sections is calculated:

[0109]

[0110] Where: Δt i Indicates the time the driver travels on this road section; R path (P) represents the overall path risk of the driver’s driving path;

[0111] Driving behavior correction factor calculation: Risky driving behavior is corrected by multiplying the current risk assessment value with a weighting factor:

[0112]

[0113] Where: α brake Indicates the weight factor of emergency braking; S brake (i, t) represents the emergency braking intensity at time t; S max Indicates the maximum value of emergency braking;

[0114] Weather Modifier:

[0115] f weather (i, t) = β1·f precip (i, t)+β2·f fog (i, t);

[0116] Where: represents the precipitation impact factor; represents the haze impact factor; P(i, t) represents the rainfall in region i at time t; V(i, t) represents the visibility in region i at time t; β1 and β2 represent weighting factors;

[0117] Road condition correction factor:

[0118]

[0119] Where: Q(i, t) represents the traffic flow of area i at time t; Q max (i) represents the historical maximum flow in area i;

[0120] Calculation of spatiotemporal cumulative risk index:

[0121] R driver (i, t) = R acc (i, t)·(1+ΔR brake (i, t)+f weather (i, t)+f trafffc (i, t));

[0122] Where: ΔR brake (i, t) represents the correction factor for sudden braking; f weather (i, t) represents the weather correction factor; f traffic (i, t) represents the traffic flow correction factor.

[0123] In this embodiment, the driver's travel trajectory is obtained through the on-board OBD data, and a spatiotemporal cumulative risk index is constructed in combination with high-risk sections, driving behavior and environmental factors. This fully considers the driver's dynamic driving behavior and real-time environmental impact, and solves the problem of traditional pricing models ignoring the driver's dynamic risk. By combining high-risk sections with specific travel trajectories, the risk status of the driver in different sections and at different times can be more accurately assessed.

[0124] Step 3: Build a graph structure with the car owner as the node and the car owner's social circle as the edge between the nodes. Use the graph convolutional network to perform convolution operations on the graph structure to learn the risk transmission effect within the social circle and generate social risk characteristics;

[0125] The graph structure is set as G = (V, E, X), where V represents a node set, each node represents a car owner; E represents an edge set, edges represent social relationships between car owners, and X represents a node feature matrix, each node contains the risk characteristics of the car owner; the node feature matrix X contains the initial risk feature vector of each car owner; the initial risk feature vector includes the overall path risk R of car owner i path (P), emergency brake correction factor ΔR brake (i, t), weather correction factor f weather (i, t) and traffic flow correction factor f traffic (i, t); where the adjacency matrix A represents the social relationship between car owners, where the value is 1 if there is a social relationship and 0 if there is no social relationship.

[0126] If a driver within a social circle habitually engages in risky driving behaviors, such as speeding, frequent sudden braking, and frequent lane changes, these behaviors not only increase their personal risk of accidents but may also impact other drivers within their social circle. This is because within certain social circles, drivers influence and imitate each other's behavior. For example, if a driver sees a friend frequently speeding without receiving a penalty, they may unintentionally imitate this behavior, increasing their own risk of accidents. Therefore, in this embodiment, by connecting drivers through social networks, this potential risk transmission mechanism between drivers can be captured, allowing for a more comprehensive assessment of each driver's indirect risk impact within their social circle.

[0127] See also Figure 2 As shown, the graph convolutional network includes a first data processing layer, a second data processing layer and a global pooling layer;

[0128] The first data processing layer includes a graph convolution layer and a multi-layer perceptron nonlinear transformation layer. The graph convolution layer performs a convolution operation on the input feature matrix to convert it into a first feature matrix. The first feature matrix is ​​used as the input of the multi-layer perceptron nonlinear transformation layer. The multi-layer perceptron nonlinear transformation layer performs a nonlinear transformation on the first feature matrix to obtain the feature matrix output by the first layer.

[0129] As a possible implementation of this embodiment, the multilayer perceptron includes two fully connected layers. The first fully connected layer receives the output of the graph convolution layer for nonlinear transformation. The output of the first fully connected layer enters the second fully connected layer. The second fully connected layer performs further nonlinear transformation based on the output of the second fully connected layer and outputs it to the second data processing layer.

[0130] In this embodiment, a multi-layer perceptron (MLP) is used to perform nonlinear transformation on the features output by graph convolution, so that the model can capture more complex social network risk patterns; the structure of the multi-layer perceptron can help the model process the features of each node more deeply, thereby improving prediction accuracy.

[0131] The second data processing layer includes a graph convolution layer and an attention mechanism layer. The graph convolution layer performs a convolution operation on the feature matrix output by the first layer to obtain a second feature matrix. The attention mechanism layer calculates attention weights based on the output second feature matrix, and obtains a feature matrix of the second layer output weighted by the attention mechanism layer based on the calculated attention weights and the second feature matrix.

[0132] In this implementation, by calculating the attention weights between nodes, we can effectively identify and strengthen the influence of key nodes in the social network (such as car owners with higher risks) on other nodes; we can allow the model to automatically pay attention to car owners with greater influence in the social circle, thereby improving the accuracy of the overall risk assessment, especially when the social circle is large or the information is complex.

[0133] Global pooling layer: Aggregates the features of all nodes in the graph structure to generate a global representation of the graph and obtain social risk characteristics. Through the global pooling layer, node features across the entire graph are aggregated to obtain a global representation; this global representation can comprehensively reflect the risk situation of the entire social circle, enabling the model to understand and assess the social risk of the driver from a global perspective.

[0134] The loss function of the graph convolutional network is as follows:

[0135]

[0136] Where: regularization and λ separation Both represent weight parameters;

[0137]

[0138] Where: N represents the number of graphs; R i represents the true social risk characteristics of the i-th graph; represents the predicted social risk characteristics of the i-th graph; represents the mean square error;

[0139]

[0140] Where: represents the global representation of the i-th graph; represents the regularization loss;

[0141]

[0142] Where: (R i -R j ) 2 Used to weight the distance between global representations; represents the global representation of the jth graph; R j represents the true social risk feature of the j-th graph; ||·||2 represents the Euclidean distance.

[0143] In this embodiment, the loss function includes regression loss, regularization loss and separation loss. Through regularization loss, the model can avoid overfitting problems and ensure the generalization ability of feature learning; separation loss helps the model better distinguish the risk characteristics of different social circles by weighting the distance between the global representations of nodes; by maximizing the differences between the social risk characteristics of different car owners, the model can optimize the risk assessment of different car owners and improve the risk differentiation between nodes in the social circle.

[0144] Step 4: Build a multi-task learning neural network based on the preprocessed data, social risk characteristics, and cumulative risk index. Use the shared network layer to learn the common features of the probability of accident and loss amount, and output the probability of accident and the predicted loss amount.

[0145] See also Figure 3 As shown, the multi-task neural network includes:

[0146] Input layer: The input vector includes social risk characteristics, spatiotemporal cumulative risk index, and preprocessed data;

[0147] Shared network layer: A multilayer perceptron is used as the shared network layer, and the shared network layer processes the input vector to obtain shared features. The multilayer perceptron uses three fully connected layers, and the input vector is processed in sequence through the three fully connected layers. The last fully connected layer outputs the shared features.

[0148] Accident probability prediction layer: The shared features are processed through the accident probability layer to obtain the probability of occurrence:

[0149]

[0150] Where: W prob b represents the weight matrix for accident probability prediction; prob F represents the bias term of the accident probability prediction layer; shared represents shared features; σ represents the sigmoid activation function; Indicates the predicted probability of danger;

[0151] Loss amount prediction layer: The loss amount prediction layer processes shared features to obtain the loss amount prediction value:

[0152]

[0153] Where: Indicates the predicted loss amount; W loss b represents the weight matrix of loss amount prediction; loss Represents the bias term of the loss amount prediction layer.

[0154] The loss function of the multi-task neural network is as follows:

[0155]

[0156] Where: risk The weight parameter representing the probability of loss; To express the loss of the probability of accident, the binary cross entropy loss function is used; Indicates the loss of the loss amount, using mean square error loss; λ loss The weight parameter representing the loss amount.

[0157] In this embodiment, by designing a multi-task learning framework, the model can share the common features of learning the probability of accident and the amount of loss; it can prompt the model to consider both the possibility of accident and the severity of loss during the learning process, rather than just processing these two tasks separately; through the shared network layer, the features of the two tasks (probability of accident and amount of loss) can leverage each other, thereby improving the overall learning effect and avoiding information fragmentation between the two.

[0158] Step 5: Generate insurance pricing based on the risk score and predicted loss amount output by the multi-task learning neural network and external market data.

[0159] The step 5 comprises the following steps:

[0160] Basic pricing: The basic pricing is determined based on the predicted probability of loss and the predicted loss amount:

[0161]

[0162] Where: Indicates the predicted probability of danger; Indicates the predicted loss amount; P base Indicates base pricing;

[0163] Base pricing adjustment: Adjust the base pricing based on market factors to obtain the adjusted insurance pricing:

[0164] P final =P base ×(1+α×ΔR+β×C market +γ×I inflation );

[0165] Where: ΔR represents the difference from the industry benchmark rate; C market Indicates market competition status; I inflation represents the current inflation rate; α, β and γ represent adjustment coefficients;

[0166] The market competition situation is based on the weighted sum of the difference from competitors' average pricing, the difference from the industry benchmark rate, the market supply and demand relationship index and the annual growth rate of market penetration rate;

[0167] Example: where the average pricing difference with competitors is calculated as follows:

[0168]

[0169] Where: P avg represents the average pricing of all competitors in the market; represents the premium pricing of the ith competitor; N represents the number of competitors;

[0170]

[0171] Where: P company represents the company's pricing; ΔP represents the difference from the average pricing of competitors.

[0172] Differences from industry benchmark rates:

[0173]

[0174] Where: R industry Indicates the standard base rate for the industry;

[0175] Market supply and demand index:

[0176]

[0177] Where: D market Indicates market demand; S market Indicates the quantity supplied;

[0178] Annual growth rate of market penetration:

[0179]

[0180] Where: P penetration，current and P penetration，previous represent the market penetration rates for the current and previous year, respectively.

[0181] In step 5, the adjustment coefficients α, β, and γ are adjusted by defining a loss function:

[0182]

[0183] Where: represents the insurance pricing after the i-th adjustment; represents the actual insurance pricing in history; the optimal adjustment coefficient is found by minimizing the loss function.

[0184] In this embodiment, basic pricing is based on the predicted probability of claims and loss amount, so that each customer's premium is highly matched with his or her actual risk, thereby improving the accuracy and fairness of pricing; secondly, the dynamic adjustment of market factors (such as industry benchmark rates, market competition conditions and inflation rates) enables insurance companies to flexibly respond to market fluctuations, adjust premiums in real time under competitive pressure and changes in the economic environment, and enhance market adaptability and competitiveness; by introducing a loss function to optimize the adjustment coefficient; the model can continuously refine pricing based on historical actual data, minimize pricing errors, and ensure that pricing is consistent with actual market conditions.

[0185] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A multi-factor integrated dynamic risk pricing optimization method for insurance products, characterized by: The following steps are involved: Step 1: Collect the car owner's insurance policy records, claims data, vehicle OBD data, meteorological data, and geographic information data, and pre-process the collected data; Step 2: Obtain the driver's travel trajectory based on the vehicle's OBD data. Based on the travel trajectory and high-risk road sections, the driver's overall road section risk is constructed. The driver's driving behavior and environmental factors are then incorporated to obtain the spatiotemporal cumulative risk index. High-risk road section calibration: Determine the accident risk value of each road section through historical data and real-time traffic monitoring: ； Where: represents the accident attenuation coefficient; represents the number of accidents on road section i at time t; represents the current timestamp; k represents the time window size; represents the accident risk value of road section i at time t; represents the base of natural logarithms; Dynamic risk path allocation: For each trip, the risk weight of the driver on different road sections is calculated: ； Where: Indicates the time the driver has traveled on the road section; Indicates the overall path risk of the driver's driving path; Driving behavior correction factor calculation: Risky driving behavior is corrected by multiplying the current risk assessment value with a weighting factor: ； Where: Indicates the weight factor of emergency braking; represents the emergency braking intensity at time t; Indicates the maximum value of emergency braking; Weather Modifier: ； Where: represents the precipitation impact factor; represents the haze impact factor; represents the rainfall in region i at time t; represents the visibility of area i at time t; and represents the weight factor; Road condition correction factor: ； Where: represents the traffic flow in area i at time t; represents the historical maximum flow of area i; Calculation of spatiotemporal cumulative risk index: ； Where: Indicates the emergency brake correction factor; represents the weather correction factor; represents the traffic flow correction factor; Step 3: Build a graph structure with the car owner as the node and the car owner's social circle as the edge between the nodes. Use the graph convolutional network to perform convolution operations on the graph structure to learn the risk transmission effect within the social circle and generate social risk characteristics; Step 4: Build a multi-task learning neural network based on the preprocessed data, social risk characteristics, and cumulative risk index. Use the shared network layer to learn the common features of the probability of accident and loss amount, and output the probability of accident and the predicted loss amount. Step 5: Generate insurance pricing based on the risk score and predicted loss amount output by the multi-task learning neural network and external market data.

2. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 1 is characterized in that: The graph structure is set to , where V represents the node set, each node represents a car owner; E represents the edge set, the edge represents the social relationship between car owners, X represents the feature matrix of the node, each node contains the risk characteristics of the car owner; The node feature matrix X contains the initial risk feature vector of each car owner; the initial risk feature vector includes the overall path risk of car owner i , emergency brake correction factor , Weather Correction Factor and traffic flow correction factors ; The adjacency matrix A represents the social relationship between car owners. If there is a social relationship, it is 1, and if there is no social relationship, it is 0.

3. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 1 is characterized in that: The graph convolutional network includes a first data processing layer, a second data processing layer and a global pooling layer; The first data processing layer includes a graph convolution layer and a multi-layer perceptron nonlinear transformation layer. The graph convolution layer performs a convolution operation on the input feature matrix to convert it into a first feature matrix. The first feature matrix is ​​used as the input of the multi-layer perceptron nonlinear transformation layer. The multi-layer perceptron nonlinear transformation layer performs a nonlinear transformation on the first feature matrix to obtain the feature matrix output by the first layer. The second data processing layer includes a graph convolution layer and an attention mechanism layer. The graph convolution layer performs a convolution operation on the feature matrix output by the first layer to obtain the second feature matrix. The attention mechanism layer calculates the attention weight based on the output second feature matrix, and obtains the feature matrix of the second layer output weighted by the attention mechanism layer based on the calculated attention weight and the second feature matrix; Global pooling layer: Aggregates the features of all nodes in the graph structure to generate a global representation of the graph and obtain social risk features.

4. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 3 is characterized in that: The loss function of the graph convolutional network is as follows: ； Where: and Both represent weight parameters; ； Where: N represents the number of graphs; represents the true social risk characteristics of the i-th graph; represents the predicted social risk characteristics of the i-th graph; represents the mean square error; ； Where: represents the global representation of the i-th graph; represents the regularization loss; ； Where: Used to weight the distance between global representations; represents the global representation of the j-th graph; represents the true social risk characteristics of the jth graph; represents the Euclidean distance.

5. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 1 is characterized in that: The multi-task neural network includes: Input layer: The input vector includes social risk characteristics, spatiotemporal cumulative risk index, and preprocessed data; Shared network layer: A multi-layer perceptron is used as the shared network layer, and the input vector is processed through the shared network layer to obtain shared features; Accident probability prediction layer: The shared features are processed through the accident probability layer to obtain the probability of occurrence: ； Where: The weight matrix representing the accident probability prediction; Represents the bias term of the accident probability prediction layer; Represents shared features; Represents the sigmoid activation function; Indicates the predicted probability of an accident; Loss amount prediction layer: The loss amount prediction layer processes shared features to obtain the loss amount prediction value: ； Where: Indicates the predicted loss amount; The weight matrix representing the loss amount prediction; Represents the bias term of the loss amount prediction layer.

6. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 5 is characterized in that: The loss function of the multi-task neural network is as follows: ； Where: The weight parameter representing the probability of loss; To express the loss of the probability of accident, the binary cross entropy loss function is used; The loss represents the amount of loss, using mean square error loss; The weight parameter representing the loss amount.

7. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 1 is characterized in that: The step 5 comprises the following steps: Basic pricing: The basic pricing is determined based on the predicted probability of loss and the predicted loss amount: ； Where: Indicates the predicted probability of an accident; Indicates the predicted loss amount; Indicates base pricing; Base pricing adjustment: Adjust the base pricing based on market factors to obtain the adjusted insurance pricing: ； Where: Indicates the difference from the industry benchmark rate; Indicates the market competition situation; Indicates the current inflation rate; 、 as well as Indicates the adjustment factor.

8. The multi-factor integrated insurance product dynamic risk pricing optimization method according to claim 7 is characterized in that: In step 5, the loss function is defined to adjust the coefficient 、 as well as Make adjustments: ； Where: represents the insurance pricing after the i-th adjustment; represents the actual insurance pricing in history; the optimal adjustment coefficient is found by minimizing the loss function.

Images