Over the years, more and more buildings got equipped with a building automation system (BAS) which provides the ability to automate building services, like heating, cooling, ventilation, air conditioning, shading, lighting, alarming, safety and security systems. Furthermore, an additional aim of a BAS is to optimize such services automatically. Such optimizations can have different purposes, e.g., to make the building more energy efficient, raise the comfort of the users, or just to make them more maintainable. For most optimization tasks, it is essential to know the future behavior of a building. Forecasts of different sensors in the building can generate this knowledge. Therefore, this thesis evaluated different algorithms to forecast such sensor data. The algorithms under investigation are ARIMA, ANN and SVM. The evaluation was divided into five categories related to different types of sensors in a building. These types of sensors were electricity, district heating, humidity, temperature, and photovoltaic production. Furthermore, the algorithms were evaluated at different seasons to ensure the accuracy of the algorithms throughout the year. This approach ended in a comprehensive performance evaluation where the ANN algorithm was superior to the others. Finally, the ANN algorithm with its elaborated network structure was implemented in a JAVA library for further development and reuse.