“…By the analysis of distinct biophysical functions in the internal interactions of dinucleotides and trinucleotides, it is possible to model nucleotides sequences as correlated structures sequences [15,19] or use these biophysical functions to model compositional features [18,20]. Since the start of the development of new computational methods for the promoter prediction research, several studies applied different classification methods, like Naive Bayes Approach [21,22], K-Nearest Neighbors (KNN) [15,19,22], Support Vector Machine (SVM) [15,19,23,22], Random Forest (RF) [19,20,24], feed-forward Artificial Neural Network (ANN) [13,25,21,22,26]. Although deep learning-based techniques achieved outstanding results, outperforming previous state-of-the-art machine learning methods for the promoter prediction problem [3].…”