@article {10.3844/jcssp.2026.1611.1619,
article_type = {journal},
title = {Animal Health Prediction Using Hybrid CNN Based BiLSTM Classification Model: A Deep Learning Approach},
author = {Rathi, J. and Sumathi, A.},
volume = {22},
number = {5},
year = {2026},
month = {May},
pages = {1611-1619},
doi = {10.3844/jcssp.2026.1611.1619},
url = {https://thescipub.com/abstract/jcssp.2026.1611.1619},
abstract = {Accurate prediction of animal diseases is vital in veterinary medicine, as it can substantially enhance animal health outcomes and mitigate economic losses, making it a critical task that warrants attention and innovative solutions. This paper proposes a novel approach to animal condition classification, leveraging an Auto-encoder-based feature selection process and an Improved Hybrid Convolutional Neural Networks (CNN) with Bidirectional (Bi-LSTM) classification methodology. The Auto-encoder-based feature selection process identifies key features in the Animal Condition Classification Dataset by learning a compressed representation and calculating feature importance scores, capturing critical information for accurate classification. The Improved Hybrid CNN with Bi-LSTM classification model combines the strengths of CNNs in feature extraction and Bi-LSTMs in sequence modeling, enabling robust classification of animal conditions. The CNN component extracts local patterns and hierarchies in the data, while the Bi-LSTM component captures long-range dependencies and contextual information. The proposed model is trained using the Adam optimizer with a categorical cross-entropy loss function and optimized through grid search, thereby demonstrating enhanced classification capabilities for animal conditions. It attains superior performance metrics including accuracy, precision, recall, and F1-score relative to existing models, thus offering a reliable and accurate solution for animal condition classification. The proposed HCNN-BiLSTM method achieved impressive results on an animal condition classification dataset, with a precision of 99.03, recall of 100, accuracy of 99.25, and F1-score of 100, outperforming CNN, LSTM, and HKNN-VNC models.},
journal = {Journal of Computer Science},
publisher = {Science Publications}
}