Development of classification model based on training time in hyperparameter for Acute Myeloid Leukemia (AML) / Nurzulaikha Zaidi@Eddie
The classification of Acute Myeloid Leukemia (AML) using machine learning models has demonstrated significant potential in advancing diagnostic accuracy and efficiency, offering critical support in clinical decision-making. This study focuses on strategies to enhance AML classification by optimizing...
Saved in:
| Main Author: | |
|---|---|
| Format: | Student Project |
| Language: | en |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/118079/1/118079.pdf https://ir.uitm.edu.my/id/eprint/118079/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The classification of Acute Myeloid Leukemia (AML) using machine learning models has demonstrated significant potential in advancing diagnostic accuracy and efficiency, offering critical support in clinical decision-making. This study focuses on strategies to enhance AML classification by optimizing hyperparameters and learning rate schedulers, aiming to reduce training time while maintaining high performance. Several learning rate schedulers, including constant, linear, step, and time-based approaches, were evaluated for their effectiveness. The results reveal that step and time-based schedulers consistently outperformed others, achieving superior accuracy, specificity, and computational efficiency, while significantly reducing training time. In addition to exploring learning rate schedulers, hyperparameter optimization techniques were applied to Convolutional Neural Networks (CNNs) such as AlexNet and ResNet-18. These techniques yielded substantial improvements in model accuracy and efficiency by fine-tuning critical parameters like learning rates and momentum. Furthermore, the study developed strategies for handling variable learning rates and momentum adjustments, with SGDM (Stochastic Gradient Descent with Momentum) showcasing excellent adaptability and convergence. This research emphasizes the importance of hyperparameter tuning and advanced optimization strategies in achieving precise and early AML diagnoses. The insights gained contribute to the development of reliable machine learning models that support personalized and effective treatment regimens, paving the way for improved clinical outcomes. |
|---|