The Effect of Principal Component Analysis on Machine Learning Accuracy with High Dimensional Spectral Data