LauraOminde01/Kenya-Road-Accident-Risk

# Kenya Road Accident Risk Dashboard A machine learning powered dashboard that analyzes Kenya road accident patterns, predicts journey risk levels and delivers evidence based road safety recommendations - built on real NTSA accident data. --- ## The Problem Kenya loses over 3,000 lives to road accidents every year. Most drivers have no way of knowing how dangerous a particular journey is before they set off. This project turns historical accident data into actionable risk intelligence. --- ## Live Dashboard > Coming soon — deploying to Streamlit Cloud --- ## What This Dashboard Does | Feature | Description | |---|---| | Accident Hotspot Map | Interactive Kenya map showing accident frequency and risk level by county | | Risk Predictor | Input your county, road and travel time to get a personalised risk score | | Analytics | Charts showing peak accident hours, dangerous roads and victim patterns | | Recommendations | Evidence based safety advice for drivers, authorities and insurers | --- ## Key Findings from the Data - **8pm is the most dangerous hour** - night driving accounts for a disproportionate share of casualties - **Sunday is the deadliest day** - weekend fatigue and late night travel elevate risk significantly - **Nairobi-Mombasa highway** has the highest recorded accident count of any road in the dataset - **85% of victims are male** - male drivers aged 25 to 40 represent the highest risk demographic - **Careless driving and speeding** are the two leading causes - both preventable --- ## Model Details The risk prediction model classifies journeys as High or Low risk based on time and location features. | | | |---|---| | Algorithm | Random Forest Classifier | | Features | Hour, night indicator, weekend indicator, peak hour indicator, county, road | | Class Imbalance | Oversampling applied to training set only | | Test Evaluation | Original unseen data - no data leakage | | High Risk Recall | 100% — model catches every dangerous journey …