This article compares the effectiveness of different drift detectors in detecting feature drift. This article compares the new Feature-Based Drift Detector (FBDD) with other state-of-the-art detectors from the literature. Because detectors cooperate with classifiers, different classifier models are also included in the research. By evaluating different drift detection methods and classification algorithms, we try to identify the best solution by combining the detector with a classifier. Both synthetic and real data sets are used in the research. The synthetic data sets simulate different types of drift (sudden, gradual, incremental, and recurring). The real data sets represent phenomena where changes in the form of drifts are unfavorable for classification quality.