... association measured by information entropy. In addition, sample weights are adjusted based on the global imbalance ratio, which enhances the penalization of majority class outliers and noisy instances. The proposed method is evaluated on 21 synthetic datasets and 35 real-world benchmark datasets with varying...