Driver Risk Scoring Using Shapley-Based Claim Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for evaluating driver safety in usage-based insurance fail to provide a holistic view of driving behaviors, unfairly penalize drivers on longer trips, and lack correlation with future claim submissions, leading to inaccurate and biased risk assessments.
Innovation Solution
A system using machine learning, specifically an extreme gradient boosting model, analyzes driver features to predict future claim frequency by calculating a risk score based on the contribution of each feature relative to others, incorporating Shapley values and spline approximations to generate a more accurate safety score.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional cumulative scoring methods are used to evaluate driver safety, then the evaluation process is simple and straightforward, but the accuracy of risk assessment deteriorates due to failure to provide holistic view and unfair penalization of longer trips
Solution Approach 1:
The patent replaces the mechanical cumulative scoring system with a machine learning-based evaluation system. Specifically, it uses gradient boosting models to process driver behavior data, automatically learning complex non-linear relationships between multiple driving events and claim outcomes, thereby achieving higher measurement precision without manual rule configuration
Solution Approach 2:
The patent transforms the evaluation approach by changing from fixed arbitrary scores to dynamic weights learned from historical data. The system adjusts the importance of different driving events based on their actual correlation with claims, allowing the evaluation parameters to adapt and optimize for accurate risk prediction
2Reliability
If arbitrary scores are assigned and subtracted for each driver action, then the scoring method is easy to implement, but the reliability of safety rating deteriorates due to lack of correlation with future claims
Solution Approach 1:
The patent implements feedback by using actual claim outcomes to train and validate the machine learning model. The system learns from historical data which driving patterns actually lead to claims, creating a closed-loop system where prediction accuracy continuously improves based on real-world results rather than theoretical assumptions
Solution Approach 2:
The patent performs preliminary action by training the gradient boosting model on extensive historical driver behavior and claim data before deployment. This pre-learning phase allows the system to establish accurate prediction capabilities in advance, ensuring reliable correlation with future claims from the start
3Ease of operation
If cumulative penalization is applied over the course of a trip, then the method provides continuous evaluation, but the fairness of assessment deteriorates by unfairly penalizing drivers on longer trips
Solution Approach 1:
The patent applies dynamics by transitioning from static arbitrary scores to dynamic context-aware evaluation. The machine learning model considers the sequence, combination, and context of driving events, adjusting the impact of each action based on the overall driving pattern and trip characteristics, thereby achieving fair assessment that adapts to different trip lengths and conditions
Data Source
AI summary
A computer-implemented system and method is provided for determining a risk assessment. The method comprises receiving a plurality of vehicle behaviour data over a defined data collection period. This data is input into a supervised learning prediction model which is trained on historical vehicle behaviour data over a past time period, to generate a predicted value of a frequency of expected claim submissions for the policyholder of the vehicle in a future time period. Then a Shapley estimate is computed for each feature of the behaviour data applied to the model for determining a contribution of each said feature to the predicted value. A spline approximation is applied to the Shapley estimate for each said feature to estimate the contribution of each said feature. Then, a sum of the spline approximation for each said feature is calculated and a corresponding risk score determined based on the sum.