A Behavioral Analysis Model and the Cause of Alcohol Dependence with the Decision Tree Technique

Abstract

Creating a behavioral model and causes of alcohol consumption by using the decision tree technique. This research aimed to study and develop behavioral analysis models about causes of alcohol use and factors affecting the cause of alcohol dependence. The sample used was students university, Freshman to Senior, Nakhon Pathom Rajabhat University, Muang District, Nakhon Pathom Province. The total number of survey takers is 500 people and the detailed data is a total of 12 characteristics. When analyzing the data, it was found that there were a total of 500 complete records available and could be classified into 4 groups: Group 1 was low risk drinkers, Group 2 was risk drinkers, Group 3 was dangerous drinkers, and Group 4 was addicted drinkers. Which uses data classification techniques with a decision tree by comparison the algorithms include the J48 algorithm, the LMT algorithm, the RandomTree algorithm, and the HoeffdingTree algorithm. By Using techniques for adjusting the balance of information with the randomly added technique. Minority samples (Synthetic minority Over-sampling Technique: SMOTE) to get the best and most accurate values to create a behavior analysis model and causes of alcohol dependence with decision tree technique. The results of this research revealed that the J48 algorithm has the data validity and precision values, which use 8-fold cross validation to divide the data.The balance SMOTE 800 has 80.61% validity value, 78.20% accuracy value, 78.80% precision value, which is more than LMT, RandomTree and HoeffdingTree.