Multi-class Model to Predict Pain on Lower Limb Intermittent Claudication Patients

Intermittent claudication is a vascular disease that hinders elder patients’ mobility, with symptoms ranging from mild discomfort to acute pain. Based on the current literature and health professional input, the most sure-fire way to diminish these symptoms is to monitor the patient’s daily walk cycle and motivate them to try to keep a steady pace despite feeling pain, the main goal being the daily increase of time the patient can walk before feeling discomfort. As such, it is of great interest to be able to predict if the patient is or will start to feel discomfort in their lower limbs, and in what specific area, based on variables such as their current speed and heart rate. By utilizing the data provided by volunteer patients throughout two months of walking tests, a few machine learning models were trained and evaluated on their success at these tasks, namely two classification tree models and two k-nearest neighbors models, one of each predicting the existence of pain and the others predicting the specific area of pain. The k-nearest neighbors models proved to be more effective at the task proposed, with a slight edge towards the pain location prediction model.