Maryam Seifaddini - Department of Computer Science (Assistant professor), Faculty of Mathematics University of Guilan, Rasht, Iran
Amir Habibdoust - Department of Economics and Accounting (Lecturer), University of Guilan, Rasht, Iran

In the area of big data, Google trends and its forecasting power have received considerable attention among scientists. In this paper, we investigate the ability of Google trend data to forecast new cases of Covide-۱۹ in Iran. We employed a supervised machine learning method known as GMDH- type neural network. The searching data of several Covid-۱۹ related terms in the Persian language are used as predictors of Covid-۱۹ new cases over the period۲۰۲۰/۱۰/۲۴ to ۲۰۲۱/۰۶/۲۲. Five models with different input variables are investigated. The results show that the RMSE of the models varies between ۱۰.۸۱ and ۸.۱۵۲, and the lowest RMSE occurs in the model in which the seven-day lag of COVID-۱۹ new cases is entered as an input variable. Our tool can improve the policymakers' and researchers’ understating from spreading pandemic during a challenging time when the infection rate is significantly high, andofficial statistics cannot be reliable

Machine Learning, GMDH type neural Network, Covid- ۱۹, Google Trends.