<p>Traditional machine learning methods often suffer from learning bias in class imbalance classification due to their over-reliance on majority class samples, leading to poor generalization performance for minority class samples. To address this issue, we propose a novel approach that integrates data and algorithmic perspectives. Our method constructs multiple feature subsets using different feature selection methods and develops an imbalanced data classification algorithm based on intuitionistic fuzzy twin support vector machine ensemble learning (EIFTSVM-CIL). First, we design an intuitionistic fuzzy twin support vector machine model based on dissimilarity measure (IFTSVM-CIL) as the base classifier, which effectively mitigates the negative impact of noise on imbalanced classification tasks. Second, multiple feature subsets are generated through various feature selection methods, and submodels are trained using IFTSVM-CIL on these subsets. Finally, we design a weighted ensemble decision-making method based on fuzzy systems, which intelligently aggregates the predictions of the submodels to produce accurate final classification results. Additionally, IFTSVM-CIL incorporates a coordinate descent strategy with shrinking by active set to reduce computational complexity, significantly enhancing the training efficiency of the model. The proposed EIFTSVM-CIL is capable of effectively addressing class imbalance learning problems, particularly in the presence of noise. Extensive numerical experiments and statistical analyses conducted on imbalanced real-world datasets from UCI and KEEL repositories demonstrate that EIFTSVM-CIL achieves superior generalization performance compared to state-of-the-art baseline models. Furthermore, we validate the practical utility of EIFTSVM-CIL through an application to cardiovascular disease detection, where it outperforms existing methods, underscoring its effectiveness in real-world scenarios.</p>