ABSTRACT
Ever since the outbreak of the novel coronavirus (2019-nCoV) epidemic in the city of Wuhan, China, which is the capital of Hubei province on December 31, 2019, the premonition of a global pandemic was always on the cards. Almost all efforts from the governments and statutory authorities have been since invested in curbing the spread of the outbreak by resorting to all possible sorts of early prediction mechanisms. Long Short-Term Memories (LSTMs) have been efficiently applied for time-series predictions of medical data-based diagnosis. In this chapter, we present a four-stacked LSTM network comprising 45 LSTM cells in each hidden stacked layer for early prediction of probable new coronavirus infections in some affected countries (India and the USA) based on real-world data sets that are analyzed using various perspectives like day-wise number of confirmed cases, test cases, number of recovered cases, and death cases, to name a few. The proposed four-stacked LSTM is experimented on the publicly available data sets from the Kaggle website, and it has been observed that the suggested model outperforms the LSTM-based state-of-the-art methods in terms of root mean square error. This attempt is to help the concerned authorities to gain some early insights into the probable devastation likely to be effected by the deadly pandemic. The python code of the proposed stacked LSTM model is also made publicly available in GitHub.
