Skip to content

Team · Team Project | Predictive Analytics

Accident Risk Prediction

A supervised machine-learning workflow that models historical road-incident data to estimate accident risk levels across environmental, temporal, and roadway conditions.

My contribution

I contributed to data preparation, feature work, model experiments, and framing outputs as risk indicators for stakeholders — not absolute predictions of individual events.

Python · Pandas · Scikit-learn · Feature engineering · Classification · Model evaluation · Imbalanced learning

Problem

Road safety interventions are often reactive due to a lack of granular risk scoring across combinations of weather, road geometry, and traffic timing.

Solution & architecture

Executed end-to-end tabular ML engineering — handling missing data, encoding categorical road/weather variables, engineering interaction features, and training classification/risk-scoring models.

Feature engineering & evaluation

Temporal bucketing (rush hour, night), weather severity encoding, and road-type interactions, with class-imbalance handling and precision-recall/ROC analysis across risk tiers.

My contribution

Team contribution across cleaning, feature engineering, experimentation, and stakeholder-facing interpretation of risk signals.

Impact

Identifies high-risk condition combinations to support preventative traffic advisories and resource allocation discussions.