Alternating decision tree applied to risk assessment of heart failure patients

About 50% of the patients diagnosed with heart failure die within four years. At the same time, a rise in home telemonitoring of these patients can be observed. For its successful deployment, predicting if a heart failure patient could die within a certain period of time is an important task. An investigation of an alternating decision tree employed for this type of prediction is presented here. Experimental results are provided showing its performance on a database which contains data about 2032 patients with heart failure. Minimizing life- threatening situations while maintaining the costs of treatment are especially targeted.