 This paper evaluates the performance of an ensemble of population and machine learning models to predict the evolution of the COVID-19 pandemic in Spain. The ensemble consists of classical ODE-based population models and machine learning models, which were adjusted to capture long-term trends. The authors then improved the machine learning models by adding additional features such as vaccination, human mobility and weather conditions. Despite the improvement, the ensemble did not perform better than the individual models, suggesting that each model has its own strengths and weaknesses. Furthermore, the models degraded when new COVID variants emerged after training, highlighting the need for continuous monitoring of the pandemic. Finally, the authors used Shapley Additive Explanation, SAX, to determine the relative importance of the various inputs for the model's predictions. This study demonstrates that an ensemble of population and machine learning models could provide a reliable and robust alternative to compartmental models, especially since they do not require data from recovered patients, which are difficult to collect and often unavailable. This article was authored by Ignacio Heredia-Casher, Judith Sainz-Pardo Diaz, Maria Castrillo, and others. We are article.tv, links in the description below.