Predicting Cancer: Leveraging Machine Learning Techniques on Clinical Data Sets

Cancer remains a leading cause of global mortality, where early detection significantly improves survival rates. This study presents a machine learning (ML) framework for cancer prediction using clinical datasets, addressing critical gaps in conventional diagnostic methods. We analyzed demographic, lifestyle, and biomarker data from 5,000 patients (breast, lung, and colorectal cancers), incorporating feature engineering to handle missing values and class imbalances. Five ML algorithms—Random Forest (RF), XG Boost, Support Vector Machines (SVM), Logistic Regression (LR), and Neural Networks (NN)—were trained to classify malignancy risk. Data preprocessing included SMOTE oversampling and Standard Scalar normalization, followed by Recursive Feature Elimination (RFE) to prioritize high-impact predictors (e.g., tumor size, genetic mutations, and biomarker levels). Hyper parameter tuning via Grid Search CV optimized model performance, evaluated using 5-fold cross-validation.