Machine Learning Based Earlier Pancreatic Cancer Detection Using Hybrid Feature Selection

Pancreatic Ductal Adenocarcinoma Cancer (PDAC), a dangerous cancer in the world, has a very poor percentage of surviving more than 5 years. Usual treatments like chemotherapy, surgery, and radiation have not been shown to significantly improve survival. If we can simplify the diagnosis of diseases through Machine Learning (ML), then the condition of Pancreatic Ductal Adenocarcinoma Cancer in humans can be known much earlier. In this research, a machine learning-based model is proposed to detect PDCA at an early stage. Different preprocessing approaches are used to increase the model’s efficiency. Chi-squared and SelectKBest are used to get the best features. Seven different ML classifiers, such as Extreme Gradient Boosting (XGBoost), LightGBM (LGBM), K-Nearest Neighbor (KNN), Decision Tree (DT), Random Forest (RF), and Extra Tree Classifier (ET), are utilized to find the best one for the prediction. Finally, it is observed that LGBM shows the best performance among others, with an accuracy of 97%.