Comparative Analysis of Machine Learning Models for Cancer Diagnosis

Pancreatic cancer is one of the most deadly cancers, with early detection being critical for improving patient outcomes. This study evaluates the performance of several machine learning models in diagnosing pancreatic cancer using a synthetic dataset. We tested models including Logistic Regression, Random Forest, Support Vector Machine (SVM), Neural Networks, Decision Trees, and SuperLearner. Despite achieving high accuracy (76.05%-76.35%), the models struggled with sensitivity, which is crucial in the context of medical diagnoses. Among the models, the SuperLearner model achieved the highest precision (66.67%), while the Random Forest failed to detect any true positive cases. This highlights the need for further improvements, such as resampling or decision threshold tuning, to enhance the sensitivity of the models. The study concludes that while more complex models like SuperLearner provide high precision, simpler models like Logistic Regression may offer a better balance between accuracy and interpretability in clinical practice.