A fundamental issue for statistical classification models in a streamingenvironment is that the joint distribution between predictor and responsevariables changes over time (a phenomenon also known as concept drifts), suchthat their classification performance deteriorates dramatically. In this paper,we first present a hierarchical hypothesis testing (HHT) framework that candetect and also adapt to various concept drift types (e.g., recurrent orirregular, gradual or abrupt), even in the presence of imbalanced data labels.A novel concept drift detector, namely Hierarchical Linear Four Rates (HLFR),is implemented under the HHT framework thereafter. By substituting awidely-acknowledged retraining scheme with an adaptive training strategy, wefurther demonstrate that the concept drift adaptation capability of HLFR can besignificantly boosted. The theoretical analysis on the Type-I and Type-IIerrors of HLFR is also performed. Experiments on both simulated and real-worlddatasets illustrate that our methods outperform state-of-the-art methods interms of detection precision, detection delay as well as the adaptabilityacross different concept drift types.