Feature Selection Method Based on Honeybee-SMOTE for Medical Data Classification
Shobha Aswal, Neelu Jyothi Ahuja, Ritika Mehra · Informatica · 2023
Bio-Medical data analysis has an important role in clinical practices. Usually, bio-medical data have complex issues like skeweedness, redundant and irrelevant attributes etc.. Several redundant and unrelated features frequently degrade the accuracy of the classifier while using with imbalanced datasets. The selection of features becomes critical in this situation. The key goal of feature selection is to establish a feature subspace that maintains classifier accuracy even as reducing the excessive computational learning cost and casting off noise. Appropriate feature selection approaches are highly dependent on their ability to match the issue context and uncover fundamental patterns within the data. This study’s main goal is to construct a disease detection model that uses a hybrid feature-selection strategy based on Honeybee-SMOTE and classification using the c4.5 algorithm. The empirical results establish the suggested hybrid methodology's superiority over competing methods regarding the accuracy parameter, precision-parameter, recall-parameter, f1-score parameter and G-Mean parameter. The statistical analysis of the collected findings demonstrates that the suggested hybrid method outperforms and is competitive with existing state-of-the-art algorithms.