Optimization of Classification Model for Early Detection of Pancreatic Cancer Using GridSearchCV and Autoencoder

Pancreatic Ductal Adenocarcinoma (PDAC) is one of the most difficult cancers to detect, often diagnosed at advanced stages. PDAC is the seventh most common cause of death worldwide. Due to malignant tumors with an extremely poor prognosis, more than 75% of patients are discovered at an advanced stage, and only 7% of patients survive for five years following diagnosis. Early detection of PDAC is crucial for improving patient survival rates. This study develops a machine learning model for the early detection of PDAC using Gradient Boosting Classifier (GBC) and Light Gradient Boosting Machine (LightGBM) algorithms with GridSearchCV optimization and Autoencoder methods to handle missing values. The dataset used is Urinary Biomarkers for Pancreatic Cancer from Kaggle. The results show that the LightGBM model, optimized with GridSearchCV, surpasses the GBC model in distinguishing patient types. The LightGBM model achieved an accuracy of 77.45<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> before optimization and improved to 78.81% after GridSearchCV optimization. This emphasizes the effectiveness of GridSearchCV in optimizing hyperparameters to improve classification performance.