Enhanced Predictions of Pancreatic Cancer Patient Survival in Local Hospital

This project focuses on enhancing precision medicine for pancreatic cancer by developing predictive models for patient survival using advanced machine learning algorithms. The research analyzed a dataset from a local hospital, consisting of 14 clinical and biomarker features, including patient demographics, diagnosis details, and biomarkers like plasma CA19-9, creatinine, LYVE1, REG1B, TFF1, and REG1A. The machine learning techniques employed include Elastic Net (EN), Decision Trees (DT), Radial Basis Function Support Vector Machine (RBF-SVM), Random Forest, and a Stacking Classifier. Extensive feature engineering was performed, incorporating Box-Cox transformation and effective handling of outliers and missing values to improve model performance. The models exhibited high accuracy in predicting patient survival, with Decision Trees, Random Forest, and the Stacking Classifier achieving perfect accuracy scores of 1.00. The Elastic Net and RBF-SVM models also performed well, with accuracy scores of 0.85 and 0.80, respectively. The study highlights the potential of these machine learning models to advance precision medicine by offering more accurate predictions of pancreatic cancer patient outcomes. This can lead to more personalized treatment plans, ultimately improving patient care and survival rates. The findings underscore the importance of integrating machine learning techniques into clinical settings to enhance the predictive accuracy of survival outcomes, thereby providing valuable insights for healthcare professionals in the treatment of pancreatic cancer.

Keywords: Precision medicine, pancreatic cancer, machine learning, Elastic Net, Decision Trees, Radial Basis Function Support Vector Machine, Random Forest, Stacking Classifier, feature engineering, Box-Cox transformation, patient survival prediction.