Categorical Modeling Method to Analyze Factors Relating to Longevity of Populations in the East and Southeast Asia

We propose in this work the categorical modeling method based on machine learning techniques to analyze environmental and economic factors anticipating to affect longevity patterns of people. The advantage of categorical modeling from the original numeric data is that the derived models are concise and easy for interpretation. We apply various learning algorithms during the modeling phase and it turns out that decision tree learning shows the best performance in classifying level of longevity according to the important factors. The classification accuracies on various countries range between 85 to 100%. The tree models also reveal prominent economic and environmental factors affecting longevity of populations living in the East and Southeast Asia regions including Japan, South Korea, Singapore, Thailand, Malaysia, Indonesia, and Vietnam. Even though the derived models differ from one country to another, there exists one common environmental factor showing negative impact to longevity in every model. This factor is the particulate emission damage, which is the loss of life due to exposure to ozone pollution and concentrations of particulates less than 2.5 microns in diameter, or PM2.5. Electric power consumption at the moderate level shows positive impact toward long life in almost every country, except Japan. The two most important environment factors appear in the longevity pattern of Japanese population are particulate emission damage and forest depletion.