Using decision tree for efficient data classification in Cyber-physical systems

Applications of sensor networks to physical systems control can benefit directly from data mining techniques. This paper presents two approaches based on decision tree for event prediction in a Cyber-physical System. Special attention was paid to efficient resources management. The fist approach is a centralized solution validated through an application of temperature control for indoor spaces. The second one is based on results of the first approach and represents a study for a decentralized solution having an innovative design that improves efficiency. Tests are conducted using Monte Carlo simulation to cover various practical situations.