Hybrid CNN and LSTM Network For Communicable Disease Prediction
Tsewang Dolker, Tene Ramakrishnudu · 2023
As reported by the World Health Organization (WHO), infectious or communicable diseases are one of the major causes of morbidity and deaths as they spread rapidly. It is crucial to diagnose these diseases early on to lessen their deadly effects. Some of the infectious diseases associated with emergencies are diarrhea, dengue, tuberculosis, COVID-19, measles, etc. COVID-19 and Tuberculosis (TB) are contagious illnesses that mainly affect the lungs. A virus, Severe acute respiratory syndrome coronavirus 2 (SARSCoV-2) causes COVID-19 and bacteria known as Mycobacterium is a reason for TB. Both of these diseases share a few similar symptoms like cough and breathing difficulty, however incubation period in TB is longer than COVID-19. There exists various diagnosis approach for COVID-19 and TB like PCR-based test, Antigen test, TB skin, and blood test but they require time, effort, and expertise. A prediction system is needed to tackle the issue, for this some of the researchers have developed various machine learning-based prediction systems but they failed to secure high accuracy. Deep learning with its recent developments has shown a noteworthy impact on data analytics and biomedical engineering (BME). In this paper a system that combines a convolutional neural network (CNN) with a long short-term memory (LSTM) network to predict COVID-19 and tuberculosis is proposed, this system achieves higher accuracy than various other traditional models. The proposed model was trained on a COVID-19 and TB symptoms dataset to classify a person as infected or not, resulting in an accuracy of 98.25% for COVID-19 and 93.22% for tuberculosis. It is evident from the results that the proposed approach outmatches other models in terms of accuracy, precision, F1-Score, and ROC Score.