Applying Decision Trees to Detect Dangerous Driving Locations

Overview

In the United States, car accidents are a leading cause of death, with tens of thousands of people dying every year. Recently, new technology has emerged that uses AI to help drivers drive safely. A notable example is Tesla’s cars, which promise self-driving and driver aids like automatic braking. This project investigated whether Decision Trees could accurately predict the severity of car accidents given data about the situation, which would give drivers a frame of reference for how dangerous the current situation is. A dataset of car accidents in the US from 2016 to 2023 was used to train multiple Decision Tree models of varying configurations and sizes. Decision Tree models are generally faster than neural networks (previously used in many studies on predicting accident severity) and also outperform them when the data is in a tabular format. Furthermore, an app was created that utilizes the decision tree model that was trained, and its performance was tested in a real-world driving situation. While the model was able to achieve accuracy rates of 78%-85% depending on the specific size, the models noticeably overfitted to the data at high sizes and were heavily influenced by uneven training data at low sizes. Furthermore, the app was able to quickly run a decision tree model to predict the possible accident severity. Future avenues for research include using the app as a platform for other models and investigating optimization methods and other types of models that could be applied.

Competition history

• AJAS 2026 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• AJAS None projects
• Every AJAS 2026 project