Exploiting fuzzy weights in CNN model-based taxonomic classification of 500-bp sequence bacterial dataset.

Taxonomic classification plays a crucial role in understanding the diversity and evolutionary relationships among bacteria. Accurately classifying bacterial DNA sequences based on a limited 500-bp segment remains challenging. This paper presents an improved Fuzzy-weighted Convolutional Neural Network (F-CNN) for taxonomic classification of bacterial DNA sequences, specifically focusing on the 500-bp segments. The proposed model aims to overcome the limitations of traditional classification methods by leveraging the power of deep learning and fuzzy logic processing. The improved fuzzy deep learning model is proposed to handle the problem of classifying samples with similar probabilities in the classification layer. It incorporates a feature selection stage using various techniques and a fuzzy weighting system to handle the uncertainty associated with similar classes in the classification layer and optimize parameters using fuzzy weights. The experimental results on the Ribosomal Database Project Release 11 (RDP 11) sequences dataset show the superiority of the proposed model, especially at the 500-bp region. Experimental results on the RDP 11 dataset, which includes over 1.4 million bacterial gene sequences, demonstrate the superior performance of the proposed model, achieving a classification accuracy up to 84.03% at the genus level for 500-bp segments and demonstrating high generalization when applied to longer sequences. This paper has significant implications for various fields, including microbiology, epidemiology, and environmental science, where accurate classification of bacteria is crucial for understanding their roles in different ecosystems and disease outbreaks.