Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Journal of Electrical Engineering and Informatics
Ari Setiawan
Universitas Bengkulu
Computer Science & IT Control & Systems Engineering Electrical & Electronics Engineering Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Performance Analysis of CNN and LSTM for SARS-CoV-2 Detection Based on Genomic Sequence Data Ari Setiawan; Ernawati Ernawati; Funny Farady Coastera
Journal of Electrical Engineering and Informatics Vol. 4 No. 1 (2026): Journal of Electrical Engineering and Informatics
Publisher : Fakultas Teknik Universitas Negeri Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59562/jeeni.v4i1.14560

Abstract

This study compares the performance of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) architecture implemented with a bidirectional configuration for binary SARS-CoV-2 detection from genomic sequence data. A total of 8,273 sequences were programmatically retrieved from NCBI, comprising 2,000 SARS-CoV-2 sequences and 6,273 sequences from other viruses. Sequences were standardized to 30,000 bases through padding or truncating and transformed using one-hot encoding. A stratified split produced 5,977 training, 1,055 validation, and 1,241 test samples. Because the experimental generator processed only complete batches of 32, final evaluation covered 1,216 of the 1,241 test samples; 25 samples that did not form a complete batch were not included in the reported metric calculations. Both models used Adam, binary cross-entropy, a maximum of 10 epochs, and EarlyStopping. CNN achieved 99.75% accuracy, 98.99% precision, 100% recall, 99.49% F1-score, and 1.0000 ROC-AUC, with 3 misclassifications. LSTM achieved 97.04% accuracy, 94.48% precision, 93.20% recall, 93.84% F1-score, and 0.9953 ROC-AUC, with 36 misclassifications. CNN required 399 seconds for training and 8 seconds for prediction, compared with 3,157 and 29 seconds for LSTM. The contribution is a controlled head-to-head comparison under a shared genomic preprocessing and training pipeline. The observed classification-efficiency advantage of CNN is specific to the dataset, implemented architectures, Google Colab/NVIDIA T4 hardware environment, and training configuration used in this experimental run and should not be interpreted as universal superiority over recurrent architectures.
Co-Authors Ernawati Ernawati funny farady Coastera