Enhanced Prediction of Severe Chemotherapy Reactions Through Comparative Analysis of XGBoost and Classical Machine Learning Models

In oncology, chemotherapy side effects remain a challenge since they could cause treatment stopping, lower quality of life, and higher mortality rates. Anticipating strong reactions early on is absolutely essential if one is to apply damage- mitigating strategies and therapeutic changes. This work evaluates five machine learning models-Decision Tree, Logistic Regression, Support Vector Machine (SVM), Random Forest, and XGboost-in terms of their accuracy in forecasting the major adverse effects resulting from chemotherapy. Early risk assessment and diagnostic accuracy should help to guide clinical decisions. To do this, the work rigorously preprocesses data, optimizes feature selection, and fine-tunes hyperparameters. Accuracy, precision, recall, F1-score define model performance. XGboost routinely beats other models in early high-risk detection. Explainability techniques based on SHAP offer interpretable and actionable model predictions to increase clinical relevance. The findings generally highlight how strong machine learning techniques could be used to develop tailored treatment plans enhancing patient outcomes and chemotherapy risk evaluation.