Relevant Feature Selection for Predicting the Severity of Motorcycle Accident in Thailand

Abstract

Thailand is the 5th ranked for road accident death toll in the world and the first ranked in Asia, which is considered a major problem in Thailand. Road accidents are an important problem that affects the quality of life of people and the country's economy. In Thailand, most of road accidents happening with motorcycles. There is the need of study for factors effecting these accidents. In this work, we applied feature selection and classification techniques for analyzing important factors causing road accidents on the dataset of motorcycle accidents. In particular, the experiment compared the performance of K-Nearest Neighbor classification models trained from (i) dataset with all features and (ii) dataset with selected features from the Wrapper technique. It was found that there was no significant difference, so the selected features could represent the models that were similar to the original ones. These selected features are the main contribution of this work since there are potential factors that can cause road accidents. This finding yield an insight information that can be incorporated as a future prevention plan for Thailand and other neighboring countries with similar environment.