DENG-Transformer: A Transformer Based Approach for Classification of Different Subtypes of Dengue Virus

Dengue virus is a significant global health issue, affecting millions of people worldwide with severe consequences and therefore timely and accurate detection and classification is a priority.We introduce a sophisticated and effective technique that uses a Transformer Neural Network to categorize the Dengue virus's four subtypes.To identify underlying patterns within the sequences, a windowing technique is first applied on 10,000 length DNA sequences, which segments the sequences into different sized chunks.After that, a Word2Vec model is applied to these segmented sequences, transforming the DNA segments into vector representations with different sizes.These vectorized representations are then passed into a Transformer based architecture together with the relevant classification labels.The model generates output probabilities for each class.Furthermore, With a vector size of 64 and a window length of 4 combined, the proposed architecture attains an astounding accuracy of 99.75%.