Enhanced Cancer Prediction Using Fine-Tuned CatBoost Model

Cancer prediction using machine learning (ML) models has gained significant attention due to its potential to improve early diagnosis and treatment outcomes. Current models often face challenges such as imbalanced datasets, inadequate feature selection methods, and suboptimal hyperparameter tuning, limiting their predictive accuracy and clinical applicability. To fill in these gaps, this study proposes a fine-tuned CatBoost model that is best for predicting cancer. It uses advanced data processing methods, such as ANOVA-based feature selection, to make sure that the input features are relevant. We benchmark the proposed study against multiple state-of-the-art ML models, including Random Forest, XGBoost, and AdaBoost, to provide a comprehensive performance comparison. By applying meticulous hyperparameter tuning, specifically optimizing parameters like learning rate, depth, and iterations, the CatBoost model significantly improves accuracy, precision, recall, and AUC scores. Results indicate a 0.93 accuracy with a 3% to 5% improvement over traditional models like decision trees and K-nearest neighbors. The CatBoost model exhibits enhanced interpretability through feature importance analysis, making it highly applicable in clinical settings. The fine-tuned CatBoost achieves a 4% increase in precision and a 6% increase in recall compared to baseline models. This demonstrates its robustness and versatility across various datasets. The findings suggest that the proposed approach bridges the identified gaps and offers a reliable, high-performing solution for cancer prediction, outperforming existing models in terms of predictive power and clinical utility.