Granular Classification for Imbalanced Datasets: A Minkowski Distance-Based Method

The problem of classification for imbalanced datasets is frequently encountered in practical applications. data to be classified this are skewed, i.e., the samples one class (the minority class) much less than those other classes majority class). When dealing with datasets, most classifiers encounter a common limitation, that is, they often obtain better performances on class. To alleviate study, fuzzy rule-based modeling approach using information granules proposed. Information granules, as some entities derived and abstracted from data, can used describe capture characteristics (distribution structure) both classes. Since geometric depend distance measures granulation process, main idea study construct each Minkowski then establish models by “If-Then” rules. experimental results involving synthetic publicly available reflect proposed distance-based method produce series shapes granular satisfying performance datasets.