Austin_Driver_Score_Predictor
Used Python Scikit-Learn to analyze Austin car crash data from 2018 to 2020 and created an interactive dashboard using a Random Forest Classifier algorithm to calculate a driver score from user features.
Overview:
The Austin Driver Score Predictor is a project aimed at analyzing motor vehicle accident data in Austin. The goal is to build a model that can predict the severity of car crashes based on various factors. This analysis can be useful for car insurance companies and consumers looking to make safer choices when purchasing vehicles.
Features:
- Utilizes the TxDot Crash Query System database, which includes details such as weather conditions, longitude, latitude, severity, and time and date of accidents from 2018-2020.
- Uses the NHTSA WebAPI to access the New Car Assessment Program - 5 Star Safety Rating, providing overall safety ratings for the cars involved in the crashes.
- Aims to answer questions about the performance of different car models in terms of accidents, the impact of weather types on accident frequency, the influence of demographics on accident frequency, and the areas in Austin where accidents occur the most.
Summary:
The Austin Driver Score Predictor project aims to analyze motor vehicle accident data to predict the severity of car crashes. By using the TxDot Crash Query System database and the NHTSA WebAPI, the project explores various factors such as car models, weather conditions, demographics, and accident locations. The analysis involves creating a PostgreSQL database, connecting to it using Psycopg2, analyzing the data with Pandas, applying machine learning algorithms, and creating a dashboard with Tableau, Javascript, Flask, HTML, and CSS. The project's initial findings suggest a low accuracy rate in predicting crash severity using linear models and neural networks, leading to the decision to use a decision tree algorithm.