A Comparison of Data Mining Techniques for Suicide Attempt Characteristics Mapping and Prediction

摘要

According to the report by Khon Kaen Rajanagarin-dra Psychiatric hospital, there are many factors such as poverty, loss, disappointment, alcohol addiction, addiction, chronic physical illness, depression, economic problems and major life changes that cause suicide attempts. However, there are some unsuccessful suicides. These individuals might try to repeatedly commit suicide attempts. In this paper, the aims of this research are to compare ID3, C4.5 and naïve Bayes and predict the characteristics of individuals who have suicidal ideation to repeatedly commit suicide attempts by using data mining techniques. After data preprocessing, three data mining techniques were selected to compare the efficiency of the single of classification models and the ensemble models. The researchers also adopted synthetic minority over-sampling technique (SMOTE) to balance the suicidal ideation of reattempt and no attempt classes. The researchers used 57 attributes from self-harm surveillance report (RP.506S) of Khon Kaen Rajanagarindra Psychiatric hospital. The experimental results of the single method showed that C4.5 is the best model with the highest percentage of correctly classified instances rate of 90.69%, while the ensemble model, voting method showed the best results of 91.26. We found the number of characteristics from C4.5 is 474 patterns. There are people, who have suicidal ideation over 500 in each case is 42 patterns. In 42 patterns, there are 22 patterns that more than 90% accuracy.