Comparative estimation of the spread of acute diarrhea and dengue in India using statistical mathematical and deep...

This study aims to forecast the spread of acute diarrhoea and dengue diseases in India by conducting a comparative analysis of statistical, mathematical (compartmental), and deep learning time series models. Utilizing weekly reported cases and fatalities from January 1, 2011, to Week 33, 2024, we evaluated ten forecasting techniques, including Regression, Bayesian Linear Regression with MultiOutputRegressor + XGBoost, SIR model, Prophet, N-BEATS, GluonTS, LSTM, Seq2Seq, and the ARIMA statistical model. Performance was assessed using mean absolute percentage error (MAPE) and root mean square error (RMSE). Our findings indicate that the ARIMA model excels in predicting acute diarrhoeal disease cases, achieving an RMSE of 317.7 and a MAPE of 2.4. Conversely, the Seq2Seq model outperforms others in forecasting dengue cases, with an RMSE of 399.1 and a MAPE of 6.3. Additionally, models such as N-BEATS and LSTM demonstrated strong predictive capabilities, while traditional models like Regression and the SIR compartmental model showed higher error rates. This research underscores the importance of selecting appropriate forecasting models to enhance disease prediction accuracy, thereby providing valuable insights for policymakers to effectively allocate healthcare resources and implement targeted intervention strategies.