Fog computing is a contemporary distributed computing concept extending from Cloud computing, which pushes the data analytics to the edge of a sensor network as far as possible. It helps avoid performance bottleneck and data analytics latency at the central server of a Cloud. However, when Fog computing is deployed, the edge nodes are responsible in data analysis including learning and recognizing patterns from the incoming data streams. Hence it is crucial to find appropriate data mining algorithm(s) which is lightweight in operation and accurate in predictive performance. In this paper, the suitability of data mining and data stream mining algorithms are investigated in Fog computing environment. Specifically, non-black-box machine learning models such as decision trees are looked into, with a quick pre-processing function implemented by correlation-based feature selection algorithm coupled with traditional search methods and particle swarm optimization search method. The simulation is based on an IoT environment where emergency services are to be supported. The results of this paper sheds light into what/which algorithms should be designed and chosen for delivering edge intelligence under Fog computing environment.