Optimization of machine learning algorithms for proteomic analysis using topsis

Abstract

The present study focuses on a new application of the TOPSIS method for the optimization of machine learning algorithms, supervised neural networks (SNN), the quick classifier (QC), and genetic algorithm (GA) for proteomic analysis. The main hypotheses are that the change in the weights of alternatives could affect the ranking of algorithms. The obtained data confirmed this hypothesis for their ranking. Moreover, adding labor as a cost criterion to the list of criteria did not affect this ranking. This was because candidate 3 had better fuzzy membership degrees than the two other candidates concerning their criteria. This work showed the importance of the value of the fuzzy membership degrees of the cost criterion of the algorithms in their ranks. The values of the fuzzy membership degrees of the algorithms used for proteomic analysis could determine their priority according to their score differences. One of the advantages of this study was that the studied methods could be compared according to their characteristics. Another advantage was that the obtained results could be related to the new ones after improving these methods. The results of this work could be applied in engineering, where the analysis of proteins would be performed with these methods.