THE ANALYSIS OF INJURED AND DEAD PEOPLE DATA FROM ACCIDENT IN SONGKRAN FESTIVAL BY USING DATA MINING TECHNIQUES

Abstract

The objectives of this research is to analysis of injured and dead people data from accident in Songkran Festival by using data mining techniques. The injured and dead people data in Songkran festival from 2008 to 2014 is classified by hospital from data.go.th is used in this research. The researcher prepared the data for classification by eliminating the unrelated data and adjusting the number of answer or result to be similarly proportional. The classification of data use the treatment results attribute is set as the desired answer. The algorithms used for create the model were Naive Bayes, Generalized Linear Model (GLM), Logistic Regression, Decision Tree, Random Forest, and Gradient Boosted Trees (XGBoost). The results showed that the most relevant attributes are male have a relationship with drinking alcohol and It was found that not wearing a helmet is the highest importance in classification. For the efficiency of the algorithms used to create model, Naive Bayes has the highest efficiency when considering on the accuracy, it has 77.4% of accuracy. Moreover, it is also found that most accidents occur on rural roads, the driver is male, not wear helmet, no parties/fall down by themselves, and usually occurs on day 13 of April. Measures to prevent death from accident should be aimed at wearing a helmet, emphasize the area at highways and city streets. Due to, if not wear a helmet, then an accident occurs on highways or city streets, based on the research results, the driver has a high probability of death.