Performance evaluation of concept drift detection techniques in the presence of noise

Many real-world applications generate massive amount of data that are continuous. This kind of data is known as streams. Sensor data, web logs, smart devices, phone records, social networks and ATM transactions are examples of sources of data streams. Typically, data streams evolve over time; this is referred to as concept drift. This phenomenon creates new challenges not present in classical machine learning techniques. In this paper, we compare 4 different methods to detect concept drift from data streams and determine their robustness in the presence of noisy data. We conducted a set of experiments on synthetic and real-world datasets. Finally, we present the results and suggest possible directions for future work.