Bi-Directional LSTM-Based COVID-19 Detection Using Clinical Reports

Abstract

COVID-19 has affected the entire globe with its rapid spreading, causing a high transmission rate. A huge amount of people come in contact with this deadly virus, and early diagnosis of such kind of viruses may save many lives. This paper proposes an improved approach for detecting COVID-19 based on Long Short Term Memory (LSTM) and taking advantage of early clinical reports. To train the LSTM-based classifier for COVID-19 detection, various preprocessing techniques and word embeddings are employed. These techniques ensure the data is in a suitable format for the LSTM model. The proposed LSTM model is then compared against state-of-the-art ensemble models like Bagging and Random Forest, demonstrating its superior performance. The evaluation results showcase a testing accuracy of 87.15%, with a precision of 91% and a recall of 88%. These metrics indicate the effectiveness of the proposed LSTM model in accurately detecting COVID-19-positive cases. By leveraging early clinical reports and utilizing advanced deep learning techniques, our approach achieves significant improvements in COVID-19 detection compared to existing ensemble models.

Keywords

• LSTM
• COVID-19
• Diagnosis
• Detecting
• Embeddings
• Classifier

References

1. Alazab, M., Awajan, A., Mesleh, A., Abraham, A., Jatana, V., & Alhyari, S. (2020). COVID-19 prediction and detection using deep learning. International Journal of Computer Information Systems and Industrial Management Applications, 12(June), 168–181.
2. Almarimi, N., Ouni, A., Bouktif, S., Mkaouer, M. W., Kula, R. G., & Saied, M. A. (2019). Web service api recommendation for automated mashup creation using multi-objective evolutionary search. Applied Soft Computing, 85, 105830.
3. Bouktif, S. & Awad, M. A. (2013). Ant colony based approach to predict stock market movement from mood collected on twitter. In Proceedings of the 2013 IEEE/ACM İnternational Conference on Advances in Social Networks Analysis and Mining, 837–845.
4. Bouktif, S., Fiaz, A., & Awad, M. (2020). Augmented textual features-based stock market prediction. IEEE Access, 8, 40269–40282.