<![CDATA[Leukemia is one of the deadliest blood cancers that urgently requires early detection for effective treatment. However, conventional diagnosis methods are often subjective, time-consuming, and expensive, posing challenges especially in resource-constrained areas. This study presents a comprehensive comparative analysis of two widely-used machine learning algorithms - Random Forest (RF) and Logistic Regression (LR) - for leukemia prediction using an open-access dataset of 10,661 preprocessed microscopic blood cell images from Kaggle. The dataset was carefully partitioned into training (80%) and testing (20%) sets, with rigorous preprocessing including image normalization and feature extraction. Our evaluation incorporated multiple performance metrics: accuracy, sensitivity, specificity, and AUC. The results show that Random Forest's performance is superior with a classification accuracy of 85.23%, specificity of 0.9351, sensitivity of 0.6774, and AUC of 0.8881, significantly outperforming LR which achieved an accuracy of 78.11%, specificity of 0.8363, sensitivity of 0.6742, and AUC of 0.8120. These findings suggest that ensemble methods like RF are particularly well-suited for detecting one of the most deadly blood cancers, leukemia, due to their ability to handle complex feature interactions in medical imaging data. While both algorithms have potential as clinical decision support, future research can test deep learning techniques and larger datasets to improve the accuracy and reliability of the model.]]>
Copyrights © 2025
