Markov and Hidden Markov Models for Genomic Sequence Classification

Abstract

The rapid growth of genomic data generated by high-throughput sequencing technologies has created significant challenges for statistical modeling and sequence classification. In this paper, we investigate genomic sequence classification using probabilistic and machine learning approaches based on Markov chains, Hidden Markov Models (HMMs), and Support Vector Machines (SVMs). Markov and Hidden Markov models are employed to capture local nucleotide dependencies and latent biological structures associated with coding and non-coding regions. Building upon these models, we introduce a hybrid HMM–SVM framework that combines generative likelihood-based features, hidden-state representations, and biologically interpretable compositional descriptors. Experimental results on genomic data demonstrate that the proposed hybrid approach substantially improves classification performance while maintaining biological interpretability.