Real-time forecasts and risk assessment of novel coronavirus (COVID-19) cases: A data-driven analysis

Tanujit Chakraborty; Indrajit Ghosh

Wavelet analysis is a mathematical tool that can reveal information within the signals in both the time and scale (frequency) domains [27] . This property overcomes the basic drawback of Fourier analysis and wavelet transforms the original signal data (especially in the time domain) into a different domain for data analysis and processing. Wavelet-based models are most suitable for nonstationary data, unlike ARIMA [23] . Most epidemic and climatic time-series datasets are nonstationary; therefore, wavelet transforms are used as a forecasting model for these datasets [11; 2] . When conducting wavelet analysis in the context of time series analysis, the selection of the optimal number of decomposition levels is vital to determine the performance of the model in the wavelet domain. The following formula for the number of decomposition levels, W L = int[log(n)] is used to select the number of decomposition levels, where n is the time-series length. The wavelet-based forecasting (WBF) model transforms the time series data by using a hybrid maximal overlap discrete wavelet transform (MODWT) algorithm with a 'haar' filter. Daubechies wavelets can produce identical events across the observed time series in so many fashions that most other time series prediction models cannot recognize [3] . The necessary steps of a wavelet-based forecasting model, defined by [2] , are as follows. Firstly, the Daubechies wavelet transformation and a 5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.