Protein ikincil yapısının tahmini için sınıflandırma yöntemlerinin optimizasyonu

Protein secondary structure prediction (PSSP) is important for understanding protein structure and function. It can be seen as a bridge between amino acid sequence and three-dimensional (3-D) structure of a protein. To date, many methods have been proposed to improve prediction accuracy. There are multiple conditions that will affect the performance of a method. One of these is the selection of correct hyper parameters, which may not be learned directly from the regular training process. Optimizing these hyper-parameters enable us to fine-tune the model complexity preventing over-fitting and under-fitting. In this thesis, we optimized a support vector machine, a deep convolutional neural field and a random forest for the second stage of a hybrid classifier for protein secondary structure prediction. In addition, we built an ensemble classifier that combines the predictions from the individual methods in various combinations. We demonstrate that the overall accuracy of the ensemble is comparable to the success rates of the state-of-the-art methods in the most difficult prediction setting and combining the selected models have the potential to further improve the accuracy of the base learners.