Ensemble classifiers with the potential of improving classification by using multiple classifiers rather than one are constructed in two steps: generating individual classifiers and combining them. Homogeneous ensemble classifiers generate an ensemble of individual classifiers by training the same learning algorithm on different training datasets and aggregate them using combining methods. In this paper, we introduce an approach called Interval-Based Ensemble Learning (IBEL) for homogeneous ensemble classifiers using interval modelling. In IBEL, bagging and information granule concepts are used to generate an interval-based output named uncertainty interval for each individual classifier in the ensemble. Then, the Interval Agreement Approach (IAA), as an interval aggregation function, is used to combine uncertainty intervals to determine the prediction of the ensemble. Uncertainty intervals capture the uncertainty of individual classifiers on their predictions, and IAA minimises information loss by capturing uncertainty between individual classifiers. In an extensive experimental study, we show the superiority of the IBEL in improving ensemble classification performance for homogeneous ensemble classifiers for both synthetic and real benchmark datasets. Furthermore, we experimentally show that in the presence of more class uncertainty, IBEL is more successful in improving ensemble classification performance than typical point prediction homogeneous ensemble classifiers.