Applications of Machine Learning in the Industry of Healthcare
The fundamentals of four traditional machine learning algorithms (DT, RF, SVM, KNN and one deep learning neural network (DNN) are introduced and how these algorithms function in assisting clinical diagnosis and disease prediction are illustrated.
Abstract
Machine learning, as a branch of Artificial Intelligence, is trying to make computers do identifications, classifications, and predictions as the way humans do, but without human involvement. Machine learning has the ability to deliver quicker and more accurate results than most traditional computer algorithms. As machine learning becomes more established, its applications are widely used. This paper is going to introduce the fundamentals of four traditional machine learning algorithms (DT, RF, SVM, KNN) and one deep learning neural network (DNN). After that, this paper will illustrate how these algorithms function in assisting clinical diagnosis and disease prediction. Final results are provided with actual experiments: DT can help practitioners identify eye diseases patients where the success rate is 92%. RF is used for diagnosing diabetes patients and it is able to achieve as high as 99.7% accuracy. By searching for similar minutiae, SVM can predict Alzheimer's patients 10 years before clinical manifestations appear, and KNN performs an 81.85% prediction accuracy for potential heart disease patients. Besides that, CNN, another form of machine learning, presents a 99% accuracy in predicting Alzheimer's patients and 83% accuracy in predicting heart disease patients.