This research focuses on the application and enhancement of machine learning algorithms for the detection and differentiation of various types of cancers, with a primary emphasis on lung cancer. Central to this study is the integration of the Bayes Optimization algorithm for hyperparameter optimization and the XGBoost algorithm for predictive modelling. A significant aspect of this work involves the strategic reduction of hyper-features, aimed at refining the XGBoost model's performance. This process not only ensures a more efficient model but also contributes to a higher accuracy in cancer-type prediction. Additionally, a comparative analysis is conducted with other ensemble models to evaluate the relative performance improvements. The findings of this study are pivotal, as they demonstrate the optimized model's enhanced capability in accurately detecting different cancer types, particularly lung cancer, and show marked advancements over other contemporary models. The research highlights the potential of combining advanced machine learning techniques for significant improvements in oncological diagnostics and treatment planning.