Online Sequential Extreme learning machine-based hybrid model for forecasting of COVID-19 cases in India

COVID-19 has posed high stress on government and people with its disruptive effects on every sector of the nation. Accurate and reliable forecasting models are of great need to handle this unprecedented situation. A hybrid model, which is a combination of, cuckoo search optimization algorithm, variational mode decomposition and online sequential extreme learning machine has been proposed in this work for multistep forecasting of COVID-19 cases. The model showed reasonable accuracy of 1.363%, 1.596% and 1.933% for one, three and five days ahead forecasting. The model gave superior results when compared with partial autocorrelation function (PACF) for selection of number of input parameters. The robustness of the proposed model has been evident in comparison with other similar state of the art techniques discussed in the literature.