This study investigates the prediction of anomalies in room climate regulated by a central Heating, Ventilation, and Air Conditioning (HVAC) system. We propose an approach to automatically annotate anomalies in unlabeled operational data from a four-story administrative building, covering 3 years of data on building energy consumption and temperature. By identifying and extracting key features, we employ various state-of-the-art machine learning methods, including Long Short-Term Memory (LSTM) neural networks, Extreme Gradient Boosting (XGBoost), and Random Forest, to predict annotated anomalies before they occur. Our results demonstrate that while all approaches perform similarly, Random Forest outperforms the others with a precision of 0.72 and a recall of 0.71. This finding provides facility managers with the ability to predict anomalies in room climate, particularly excessive heating or cooling, thus enabling cost reduction and improved operational efficiency.