Principal Component Analysis Coupled with Artificial Neural Networks for Therapeutic Indication Prediction of Thai Herbal Formulae

Abstract

This study illustrated the principal component analysis coupled with artificial neural networks (PC-ANN) as a useful tool in therapeutic indication prediction of Thai herbal formulae official in the National List of Essential Medicine 2011 and the National Traditional Household Remedies.  A set of 71 herbal formulae from the National List of Essential Medicine 2011 and the National Traditional Household Remedies and 19 formulae without therapeutic indication was used as a training set, a monitoring set and a validation set.  The performance of the model was measured by the percentage of “correctly classified”, True Positive rate and False Positive rate of the PC-ANN model.  The results suggested that principal component analysis technique could condense all of the variables in which there were interrelated, into a few principal components, while retaining as much variation presented in the data set as possible.  The use of a PC-ANN technique provided a good prediction of therapeutic indication of these herbal formulae as well as distinguishing these formulae from the one without therapeutic indication.