Overview

Road accident risk prediction is an important application of machine learning in transportation and road safety. By analyzing factors such as road type, weather conditions, lighting, and time of day, predictive models can help identify road segments with a higher likelihood of accidents. These insights can support proactive accident prevention, safer road management, and data-driven transportation planning.

In this challenge, your task is to build a machine learning regression model that predicts the likelihood of road accidents for each road segment. You will work with a structured dataset containing various road and environmental features. This challenge provides hands-on experience with the complete machine learning workflow, including data preprocessing, exploratory data analysis, feature engineering, model training, and evaluation for regression.

Practice Skills

In this challenge, you will gain hands-on experience with:

• Python
• Data preprocessing
• Exploratory Data Analysis (EDA)
• Feature engineering
• Regression algorithms

Evaluation

Goal

Train a regression model that predicts the likelihood of accidents for each road segment in the test set. The predicted risk should be a continuous value between 0 and 1.

Metric

Submissions are evaluated using Root Mean Squared Error (RMSE) between the predicted and actual accident risk values.

$\text{RMSE} = \sqrt{\frac{1}{N}\sum_{i=1}^{N}(y_i-\hat{y}_i)^2}$

Where:

• $N$ is the total number of road segments.
• $y_i$ is the actual accident risk.
• $\hat{y}_i$ is the predicted accident risk.

A lower RMSE indicates that the predictions are closer to the actual values.

Submission

Format

A CSV with two columns:

• id: The unique identifier for each road segment.
• accident_risk: The predicted likelihood of accidents, represented as a continuous value between 0 and 1.

id	accident_risk
1001	0.352
1002	0.587
1003	0.102

See the submission guide for upload instructions.

Tracks

The challenge accepts two submission tracks:

• Public: predict on the public test set and upload the resulting CSV.
• Private: upload your model. It is scored on a held-out private test set. Your code must recursively scan the entire data directory.