Forecasting COVID-19 daily cases using phone call data

The need to forecast COVID-19 related variables continues be pressing as the epidemic unfolds. Different efforts have been made, with compartmental models in epidemiology and statistical such AutoRegressive Integrated Moving Average (ARIMA), Exponential Smoothing (ETS) or computing intelligence models. These proved useful some instances by allowing decision makers distinguish different scenarios during emergency, but their accuracy has disappointing, forecasts ignore uncertainties less attention is given local areas. In this study, we propose a simple Multiple Linear Regression model, optimised use phone call data number of daily confirmed cases. Moreover, produce probabilistic that allows better deal risk. Our proposed approach outperforms ARIMA, ETS, Seasonal Naive, Prophet regression model without data, evaluated three point error metrics, one prediction interval two measures. simplicity, interpretability reliability obtained careful forecasting exercise, meaningful contribution at level who acutely organise resources already strained health services. We hope would serve building block other on hand help front-line personal level, facilitate communication modelling being made national improve way tackle pandemic similar future challenges.