Combining covid-19 case prediction and analysis of seasonal data impacts using deep learning methods

The new coronavirus (Covid-19) epidemic, which has affected the whole world, has infected millions of people and caused the death of hundreds of thousands, even millions of people. This epidemic has greatly affected the economics and social life of many countries. The measures taken could not prevent this and the society was caught unprepared. Estimating the rate of increase in the number of cases, is of great importance especially in the planning of administrative processes related to health infrastructure. Mathematical models and deep learning methods are used for these predictions. It is also being developed in various artificial intelligence-based approaches. In this study, in order to predict the changes in the number of COVID-19 cases in Italy, a forward forecast is made with a long short-term memory (LSTM) based neural network approach. In the study, the number of daily cases, deaths and recovered patients in Italy between Feb 24 and Nov 1, 2020 were used. In addition to, the effects of seasonal changes on the epidemic are analyzed using the meteorological data of this period. In addition to this, considering the 14-day incubation period in the COVID19 outbreak, the effect of historical values of meteorological parameters on cases is demonstrated by experimental studies. The results show that the Long short-term memory (LSTM) method can provide a significant advantage in case prediction to take preventive steps, and at the same time, seasonal data are added to the LSTM network, increasing the success rates in case prediction. Keywords: Deep learning, Artificial Neural Networks, Long Short-Term Memory (LSTMs), Pandemic, COVID-19