Clinical Data Driven Machine Learning Model for Breast Cancer Prediction
S Kanagamalliga, Srigitha S. Nath, R. Latha, Lakshmi Prabha J, Usha Rani D I, N G Praveena · 2025
This research explores the enhancement of breast cancer diagnosis using advanced Machine learning (ML) techniques, with LightGBM being selected as the primary algorithm due to its efficiency and accuracy in handling large datasets. The collected data underwent rigorous preprocessing, where categorical variables were encoded, features were normalized to eliminate biases caused by varying magnitudes, and inconsistencies were addressed to ensure data reliability. The dataset was split into testing and training subsets to maintain balance and support robust evaluation. LightGBM was implemented with default parameters initially, followed by hyperparameter tuning to optimize its performance, enabling faster training without compromising predictive accuracy. Clinical features were systematically input into the model, and predictive analysis was carried out to classify cases as benign or malignant. The trained model was evaluated using accuracy, precision, recall, and cross-validation techniques to confirm its robustness across various data partitions. Errors observed during testing were rectified through further refinements, ensuring the model’s reliability. Comparisons with other ML algorithms demonstrated LightGBM’s superior diagnostic performance, in terms of speed and accuracy. Statistical analysis provided deeper insights into the model’s strengths and limitations, highlighting its potential for real-world applications. The research highlights the critical role of preprocessing, validation, and model tuning in achieving reliable outcomes. The findings reveal that LightGBM can serve as a powerful and scalable tool for breast cancer prediction, paving the way for future advancements in healthcare diagnostics and contributing to improved patient outcomes.