Forecasting Covid-19 Cases in Türkiye With the Help of Lstm

Abstract

Even though, it is thought that the pandemic has come to an end, the humanity is still under the danger of upcoming pandemics. In that sense, every effort to understand or predict the nature of an infectious disease is very precious since those efforts will provide experience for upcoming infectious disease epidemic/pandemic. Mathematical models provide a common way to analyze the nature of the pandemic. Apart from those mathematical models that mostly determine which variables should be used in the model to predict the nature of the epidemic and at which rate the disease will spread, deep learning models can also provide a fast and practical tool. Moreover, they can shed a light on which variables should be taken into account in the construction of a mathematical model. And also, deep learning methods give rapid results in the robust forecasting trends of the number of new patients that a country will deal with. In this work, a deep learning model that forecasts time series data using a long short-term memory (LSTM) network is used. The time series data used in this project is COVID-19 data taken from the Health Ministry of Republic of Türkiye. The weekend isolation and vaccination are not considered in the deep learning model. It is seen that even though the graph is consistent and similar to the graph of real number of patients, and LSTM is an effective tool to forecast new cases, those parameters, isolation and vaccination, must be taken into account in the construction of mathematical models and also in deep learning models as well.