COMPUTATIONAL EVALUATION OF DIFFERENT MARKER PROTEINS USING MACHINE LEARNING
In this paper we have discussed the mean (observed and predicted), standard deviation, sum of squared error, absolute. mean error using different classifiers like k-NN and SVM of seven marker proteins (AkT, EGFR, ERK, IRS, MK2, JNK and FKHR) which occur due to the combination of TNF, EGF and Insulin. For k-NN we have used three different methods i.e. Chebyshev, Cityblock and Euclidean while for SVM we have used linear, polynomial, RBF and Sigmoid method by Type 1 and Type 2 approaches. Results using Euclidean method of k-NN classifier and RBF method of SVM classifier were good. In this paper we have also discussed the training, test & validation perfection, training algorithm, hidden & output activation function using different approaches of ANN of different marker proteins. Different training algorithms like BFGS and RBFT were used. We have used different types of activation functions like gaussian, exponential, logistic, tanh etc which was used as hidden activation and output activation. Results are the best in all cases for MLP instead of RBF.