Paper: arXiv 2603.15839

Authors: Jongtaek Lee, Andrei Badescu, X. Sheldon Lin

Abstract

In this paper, we propose a novel frequency-severity joint trip-level risk index that combines the frequency of abnormal driving patterns with a severity component reflecting how extreme such behavior is relative to a portfolio-level baseline. Severity is quantified through an inverse-probability penalty that increases with the rarity of observed tail extremes, rather than being interpreted as a claim size. Based on high-frequency telematics data, we construct a multi-scale representation of longitudinal acceleration using the maximal overlap discrete wavelet transform (MODWT), which preserves localized driving patterns across multiple time scales. To capture severity as tail rarity, we model the portfolio distribution using a Gaussian-Uniform mixture with a layered tail structure, where Gaussian components describe typical driving behavior and the tail is partitioned into ordered severity layers that reflect increasing extremeness. We develop a likelihood-based estimation procedure that makes inference feasible for this mixture model. The resulting severity layers are then used to construct multi-layer tail counts (MLTC) at the trip level, which are modeled within a Poisson-Gamma framework to yield a closed-form posterior risk index that jointly reflects frequency and severity. This conjugate structure naturally supports sequential updating, enabling the construction of dynamically evolving driver-level risk profiles. Using the UAH-DriveSet controlled dataset, we demonstrate that the proposed index enables reliable discrimination across behavioral driving states, identification of high-risk trips, and coherent ranking of drivers, yielding a purely behavior-driven risk measure suitable for actuarial ratemaking and potentially mitigating fairness concerns associated with traditional covariates.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a mathematically sophisticated framework combining MODWT multi-scale signal processing with a Gaussian-Uniform mixture model and Poisson-Gamma Bayesian updating for telematics risk assessment. While the empirical evaluation relies on the controlled UAH-DriveSet dataset rather than a massive commercial production backtest, the statistical methodology and novel approach to frequency-severity modeling are well-developed and clear.

Research Flowchart

  flowchart TD
    A["<b>Research Goal</b><br>Create a frequency-severity trip & driver risk index using telematics"] --> B["<b>Input Data</b><br>High-frequency longitudinal acceleration (UAH-DriveSet)"]
    
    B --> C["<b>Multi-Scale Decomposition</b><br>Extract localized driving patterns via MODWT"]
    
    C --> D["<b>Tail Rarity Modeling</b><br>Fit Gaussian-Uniform Mixture to capture severity via tail layers"]
    
    D --> E["<b>Trip & Driver Risk Index</b><br>Calculate Multi-Layer Tail Counts & apply Poisson-Gamma conjugate update"]
    
    E --> F["<b>Key Outcomes</b><br>Dynamic risk profiles, accurate driver ranking & behavior-driven ratemaking"]

Cited References (48)

  1. A Longitudinal Analysis of the Impact of Distance Driven on the Risk of Car Insurance Policyholders (2020) — link
  2. Driver Risk Identification Using Telematics and Contextual Data Crossed with Claim History (2023)
  3. Improving Automobile Insurance Claims Frequency Prediction with Telematics Car Driving Data (2022) — link
  4. Functional Data Analysis for Sparse Longitudinal Data (2005) — link
  5. Functional Boxplots (2011) — link
  6. Assessing Driving Risk Through Unsupervised Detection of Anomalies in Telematics Time Series Data (2025)
  7. Need Data for Driver Behaviour Analysis? Presenting the Public UAH-DriveSet (2016) — link
  8. On the Generalised Distance in Statistics (1936)
and 40 more
  1. Estimation and computations for Gaussian mixtures with uniform noise under separation constraints (2022) — link
  2. Wavelet Methods for Time Series Analysis (2013)
  3. Maximum Likelihood Estimation of Heterogeneous Mixtures of Gaussian and Uniform Distributions (2011) — link
  4. Estimating the Components of a Mixture of Normal Distributions (1969) — link
  5. Strong Consistency of the MLE for Finite Location-Scale Mixtures When the Scale Parameters Are Exponentially Small (2006) — link
  6. Trimmed $k$-Means: An Attempt to Robustify Quantizers (1997) — link
  7. Network-level traffic flow prediction: Functional time series vs. functional neural network approach (2024) — link
  8. Improving Many Volatility Forecasts Using Cross-Sectional Volatility Clusters (2020) — link
  9. Analysis of Accident Risks from Driving Behaviors (2017) — link
  10. Unfallrekonstruktion – Anfahrbeschleunigungen (2006)
  11. Covariate selection from telematics car driving data (2017) — link
  12. Feature extraction from telematics car driving heatmaps (2018) — link
  13. Claims frequency modeling using telematics car driving data (2019) — link
  14. Multivariate credibility modelling for usage-based motor insurance pricing with behavioural data (2019) — link
  15. Can Automobile Insurance Telematics Predict the Risk of Near-Miss Events? (2020) — link
  16. Clustering driving styles via image processing (2021) — link
  17. What can we learn from telematics car driving data: A survey (2022) — link
  18. The added value of dynamically updating motor insurance prices with telematics collected driving behavior data (2022) — link
  19. How Much Telematics Information Do Insurers Need for Claim Classification? (2022)
  20. Combined Actuarial Neural Networks for Claim Count Modeling with Mixed Features: An Application to Telematics Insurance (2024) — link
  21. Auto Insurance Pricing Using Telematics Data: Application of a Hidden Markov Model (2024)
  22. Weekly dynamic motor insurance ratemaking with a telematics signals bonus–malus score (2025) — link
  23. Data Mining of Telematics Data: Unveiling the Hidden Patterns in Driving Behaviour (2025)
  24. Decision-adjusted Driver Risk Predictive Models Using Kinematics Information (2021) — link
  25. Smartphone-based Hard-braking Event Detection at Scale for Road Safety Services (2022) — link
  26. Using telematics data to find risky driver behaviour (2019) — link
  27. Empirical Bayes: Concepts and Methods (2021)
  28. Robust Hyperparameter Estimation Protects against Hypervariable Genes and Improves Power to Detect Differential Expression (2016) — link
  29. Time and distance to first accident and driving patterns of young drivers with pay-as-you-drive insurance (2014) — link
  30. Telematics and Gender Discrimination: Some Usage-Based Evidence on Whether Men’s Risk of Accidents Differs from Women’s (2016) — link
  31. Exposure as Duration and Distance in Telematics Motor Insurance Using Generalized Additive Models (2017) — link
  32. Improving Automobile Insurance Ratemaking Using Telematics: Incorporating Mileage and Driver Behaviour Data (2019) — link
  33. The Use of Telematics Devices to Improve Automobile Insurance Rates (2019) — link
  34. Bibliometric Review of Telematics-Based Automobile Insurance: Mapping the Landscape of Research and Knowledge (2024) — link
  35. Balancing Risk Assessment and Social Fairness: An Auto Telematics Case Study (2024)
  36. The value of vehicle telematics data in insurance risk selection processes (2017)
  37. Unravelling the predictive power of telematics data in car insurance pricing (2018) — link
  38. Abstract Inference (1981)
  39. Nonparametric maximum likelihood estimation by the method of sieves (1982)
  40. Order Restricted Statistical Inference (1988)

0 of 48 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.