Predict Driving Risk

Using Machine Learning to Predict the Relative Risk of Driving Using Weather Conditions and Driver Statistics

Abstract: Every year, there are over 1.2 million accidents alone attributed to weather. Out of these, there are around 6,000 deaths and over 445,000 injuries. This project is about predicting the risk of driving on any given day using weather conditions, road conditions, and driver statistics. This model would take inputs such as the driver’s age, past driving history, current and previous weather state, wind speed, visibility, precipitation, etc. Using these inputs, we would train a neural network to output a number between 0 and 1 representing the relative risk of driving under those inputted conditions (with 0 being the lowest possible risk and 1 being the highest possible risk). Our goal is to train a model that can, with at least 95% accuracy, predict the risk of driving under certain conditions with 0.03 error. The applications of this project include using it to fine-tune self-driving cars based on the weather conditions (icy roads require more cautious driving, warm days not as much), and by using weather forecasts, our model can predict future risks as well. This will help to overall decrease the number of accidents that occur on the road.


AI Camp Project Pitch.mp4

Amy Chen and Kaavya Thirumurugan

Accolade



Top Project Idea Winner - AI Camp, Summer 2020

by Community AI Inc.