One of the most contestable problems in online learning is concept drift. In addition, if the data stream has imbalanced data, the detection of concept drift is more difficult, especially, when drift is in minority samples. Ensemble classifiers are also effective for the data stream classification with concept drift. By adjusting the weight to every individual classifier, we can manage the concept drift and misclassification problems. Using association rule mining techniques can help in balancing datasets and detecting concept drift in the early levels. In this article, we propose an Ensemble Fuzzy association Rule-based Classifier for Imbalanced data with Concept drift (EFR-IC) to deal with imbalanced streaming data containing concept drift. EFR-IC has five advantages compared with the existing methods as follows: 1) it does not need the data from previous chunks so in terms of storage space is more economical than similar methods; 2) it is stable in stationary and nonstationary environments; 3) due to the synchronization of all steps of algorithm execution -handling imbalanced data, concept drift detection, classification- execution speed is much better than similar methods; 4) it can be adapted to the new condition when swapping majority class to minority class; 5) it can timely react to multiple kinds of concept drifts. Experiments on both real and synthetic datasets containing concept drift show the effectiveness of EFR-IC in learning nonstationary imbalanced data sets.