An Alternative Approach to Causes of Death Prediction Using Support Vector Machines

The analysis of the causes of death provides valuable insight into changes in the overall mortality trends hidden in population-level data. During the past two centuries, the causes of death have shifted from infectious diseases to chronic diseases, and, consequently, the mortality reduction has moved from childhood to old age. In addition, the evolution of mortality by causes, monitored only for a small set of countries, may present different classifications over time due to regular revisions of classification rules and consequent disruption in long-term cause-of-death analysis. We develop and apply a novel statistical framework leveraging a machine learning algorithm to deal with cause-specific mortality. Our approach consists of using Support Vector Machines to classify mortality rates into specific macro classes of death causes. The proposed model is formulated by exploiting the features of mortality dynamics such as age, gender, and central death rate and could be used to classify data for countries where causes of death are misreported. We consider internationally-classified cause-of-death categories and data obtained from the Human Cause-of-Death Database for all the available countries.