"AN EFFICIENT TUBERCULOSIS INCIDENCE RATE PREDICTION USING THE ARIMA-NARX HYBRID ALGORITHM WITH FEATURE EXTRACTION"

Despite advances in healthcare, tuberculosis (TB) remains a global health concern, requiring effective planning and resource allocation for its control and prevention. Because machine learning algorithms can make predictions for the future by analyzing data, they are used in many areas and provide effective results. However, there are insufficient studies in the related literature that include European countries such as Australia, Belgium, Bulgaria, and Cyprus; therefore, more research is needed to predict the factors that trigger tuberculosis. For this reason, this research presents a novel approach for predicting Tuberculosis (TB) incidence rates using a hybrid algorithm, combining “AutoRegressive Integrated Moving Average” (ARIMA) and “Nonlinear AutoRegressive with eXogenous inputs” (NARX) models. To enhance prediction accuracy, a feature extraction process was employed, incorporating relevant variables. A public dataset of TB incidence rates (1995–2020) obtained from the World Health Organization (WHO) was used to train and test the proposed hybrid method. The performance of the proposed method was evaluated using standard metrics (MSE, RMSE, MAE, and MAPE). The results obtained indicate a significant prediction accuracy of the proposed model compared with the existing methods with overall best predictive average accuracies: ARIMA MSE: 0.6747; RMSE: 0.5227; MAE: 0.5631; and MAPE: 0.8896, while of MSE: 0.1160; NARX with MSE: 0.2092; RMSE: 0.3207; MAE: 0.2836; MAPE: 0.0865, and the proposed method (ARIMA-NARX) RMSE: 0.1566; MAE: 0.1533, and MAPE: 0.0239. We hope that this research contributes to the field of epidemiology and disease prediction and will assist healthcare organizations and policymakers in allocating resources effectively for diagnosis, treatment, and TB preventive measures.