Optimized machine learning based comparative analysis of predictive models for classification of kidney tumors

Area of Science:

• Nephrology
• Medical Imaging
• Artificial Intelligence

Background:

• Kidney tumors pose significant health risks.
• Early detection and classification are crucial for effective treatment and patient outcomes.
• Machine learning offers potential for improving diagnostic accuracy.

Purpose of the Study:

• To evaluate the efficacy of various machine learning models in detecting and classifying kidney tumors.
• To identify the optimal model and hyperparameters for kidney tumor classification.

Main Methods:

• Utilized Decision Tree, XGBoost Classifier, K-Nearest Neighbors (KNN), Random Forest, and Support Vector Machine (SVM) models.
• Dataset splitting ratios of 80:20 and 75:25 were employed, with 80:20 yielding superior performance.
• Optimized models using batch sizes of 16 and 32, and Adam and SGD optimizers.

Main Results:

• The 80:20 data split improved model performance.
• Support Vector Machine (SVM) demonstrated the highest accuracy at 98.5% with a batch size of 32 and Adam optimizer.
• K-Nearest Neighbors (KNN) achieved 90.4% accuracy.

Conclusions:

• Machine learning models, particularly SVM, show high potential for accurate kidney tumor detection and classification.
• This approach can assist healthcare professionals in early disease diagnosis.
• Further research can refine these models for clinical application.