The COVID-19 pandemic has exposed significant weaknesses in healthcare systems, especially in hospital resource management in countries with limited capacity, such as Kosovo. This study proposes a new hybrid model for predicting the spread of infection during a pandemic and the optimal use of hospital capacities. It combines the mathematical model SEIR with the machine learning technique k-Nearest Neighbors (kNN). The model parameters are optimised using the gradient descent method. The results of using the model are compared with available real-world data for hospitals in eight municipalities in Kosovo. These results validate the performance of the proposed model. This model can be used in real-world situations in future pandemics to predict the development of the pandemic and support decision-making to improve the efficiency of hospital resource utilization. The goal is to increase the resilience and preparedness of the Kosovo healthcare system for future pandemics.