Road Accident Severity Classification in Thailand Using K-Means Clustering and Decision Tree Techniques

Abstract

This research aims to analyze road accident patterns in Thailand and identify factors influencing accident severity using data mining techniques, including K-Means clustering and Decision Tree classification. The dataset used in this study consists of 81,744 accident records obtained from the Open Government Data of Thailand.

The results of K-Means clustering indicate that the optimal number of clusters is four (k = 4), which effectively categorizes road accidents into four major groups: (1) Multi-vehicle accidents with moderate to high severity, (2) Minor accidents with low severity, (3) Fatal accidents with high severity, and (4) Injury-related accidents with moderate severity. Subsequently, a Decision Tree model was employed to extract decision rules (If–Then rules), where accident severity was defined as the target variable, while temporal factors, road characteristics, vehicle types, and environmental conditions were used as independent variables.

Model performance was evaluated using 10-fold cross-validation, which achieved an Accuracy of 85%, Precision of 85%, Recall of 85%, and an F1-score of 84%, indicating that the model is capable of effectively explaining road accident patterns. The findings of this study can serve as empirical evidence to support road safety policy formulation and the development of preventive measures by relevant authorities. Furthermore, this research provides guidelines for the future application of machine learning and deep learning techniques to enhance road safety management and accident prevention strategies.