Machine Learning-Based Prediction of COVID-19 Cases in 75 Countries Using Google Trends: A Retrospective Study

Few studies have systematically examined the predictive capacity of digital search data for forecasting COVID-19 cases on a global scale, particularly in terms of search queries in different languages. Additionally, limited research has explored the correlations and temporal dynamics between digital search data in different languages and clinical cases. Therefore, this study employed Google Trends using COVID-19-related search terms in English and local languages to predict World Health Organization reported cases across 75 countries, spanning Africa, the Americas, Asia and Oceania, and Europe. Correlations between Google Trends and clinical case data were evaluated, and time-lagged cross-correlation analysis was employed to estimate the temporal dynamics of search data relative to cases. We also developed a machine learning-based random forest model to predict clinical cases using combined English and local language Google Trends. Our results identified leading time lags in most countries, suggesting that Google Trends can provide early warnings of reported clinical case surges. Furthermore, our random forest model successfully predicted COVID-19 cases using combined English and local language search data in a retrospective analysis, with identified relative contributions of English versus local-language searches. This study demonstrates that digital search data provide an openly accessible source of information that may assist forecasting of infectious disease outbreak surges prior to clinical data reporting. Digital epidemiology-based approaches could support public health decision-making by potentially complementing existing clinical and wastewater surveillance systems for infectious diseases. This work directly contributes to the United Nations sustainable development goals (SDGs), particularly SDG 3 (Good Health and Well-Being) and SDG 11 (Sustainable and Resilient Cities and Communities).