A Comparison of SVM and ELM Algorithms Based on SMOTE for Anemia Classification Using Hematology Data

Abstract

Anemia remains a significant global health concern, and its diagnosis through manual interpretation of Complete Blood Count (CBC) results is susceptible to bias and misinterpretation. Machine learning techniques offer a promising solution for identifying complex patterns in medical data. however, their performance is often affected by class imbalance issues commonly found in healthcare datasets. Therefore, this study aims to evaluate and compare the performance of Support Vector Machine (SVM) and Extreme Learning Machine (ELM) algorithms enhanced with the Synthetic Minority Over-sampling Technique (SMOTE) for anemia classification. The proposed approach employs SVM and ELM classifiers with parameter optimization using K-Fold Cross Validation, while SMOTE is applied to address the imbalance in class distribution. The study utilizes a secondary CBC dataset consisting of 364 patient records categorized into Anemia and Non-Anemia classes. Experimental results indicate that the SMOTE-based SVM model achieved an accuracy of 94.52%, precision of 97.14%, recall of 91.89%, and an F1-score of 94.44%, with a computation time of 0.013 seconds. In comparison, the SMOTE-based ELM model attained an accuracy of 91.78%, precision of 89.74%, recall of 94.59%, and an F1-score of 92.11%, while requiring only 0.002 seconds of computation time. The findings suggest that SVM delivers more stable performance and the highest precision, making it highly effective in reducing false positive predictions. On the other hand, ELM demonstrates greater sensitivity to the incorporation of synthetic samples but outperforms SVM in terms of recall and computational efficiency, making it a suitable alternative when rapid processing and higher sensitivity are prioritized.