A Fast Parkinson's Disease Prediction Technique using PCA and Artificial Neural Network

Disease Diagnosis in early stages is crucial in today's world. Big data has been a boon in the medical and health care industry. There have been various advancements in machine learning algorithms that have led to the efficient prediction of diseases. However, the high dimensionality of the data makes it more challenging and complex data. The number of features can be reduced by applying a feature selection method. But achieving the same or better accuracy on data with reduced features is a more challenging task. Therefore, the feature should be selected carefully to achieve better detection. In this paper, we propose a disease prediction approach that implements principal component analysis(PCA) to select the best suitable features from the data. The data with selected features are applied to a neural network prediction model based on back propagation. We observed a relative improvement in the detection accuracy up to 97% after selecting the most meaningful features from 754 features. We also studied the effect of model parameters on detection accuracy.