Reproducibility package for: Forecasting Reconstructed SIR Transmission Rates for Short-Term COVID-19 Prediction in Bangladesh: Neural Networks versus Linear Autoregression

Reproducibility package for the article "Forecasting Reconstructed SIR Transmission Rates for Short-Term COVID-19 Prediction in Bangladesh: Neural Networks versus Linear Autoregression" (M. B. Hossain, M. S. Alam, Molla, A.A The package contains the compiled daily COVID-19 record for Bangladesh (3 April 2020 to 8 January 2022; tests, confirmed cases, deaths) as the source PDF and an extracted CSV; Python code for causal smoothing, SIR reconstruction, the neural, linear, constant-rate, persistence, direct-autoregression and seasonal-naive forecasters, rolling-origin evaluation, statistical tests (HAC, Holm, block bootstrap, HLN) and figures; the full forecast and origin-level metric outputs; and a verification report showing that 178 selected published values are matched at the reported precision. See README.md for instructions.