Classifying Car Crash Types to Enhance Airbag Performance: A Machine Learning Approach

Cars are heavily used in life, and therefore, the safety of passengers is a priority. Since the invention, the importance of airbags as a safety feature has always been used until today. Research indicates that they help to reduce the risk of severe injuries and fatalities during accidents. Therefore, they must be used in vehicles. This paper classifies car crash types for improved airbag performance as a measure to enhance passenger safety. The research is based on data from the National Highway Traffic Safety Administration (NHTSA) company. The company provides multidimensional acceleration data and makes it available for public investigation on its website. Data will be analyzed and used to categorize of six different types of car accidents: frontal collision, oblique collision, rear collision, pole collision, offset collision, and rollover. The research covers 45 different vehicles including Honda, Toyota, Jeep, Genesis, Lexus, etc. This data makes a classification of the crash types possible and allows for differentiation among the 45 vehicles. We can also classify car types. Comparative analysis will be done based on the performance of machine learning algorithms and the results will be documented. Six machine learning algorithms are being compared, i.e., Decision Tree (DT), Adaboost Classifier, Random Forest (RF) Classifier, Multilayer Perceptron (MLP), Radial Basis Function (RBF), K-Nearest Neighbors (KNN) to find a model to distinguish different kinds of vehicles and different types of car crashes. From the comparison of DT, Adaboost, and RF algorithms with the other three algorithms, we have better accuracy, and we reached a more precise and accurate result of 100%.