Classification of Pancreatic Cancer using Kolmogorov-Arnold Networks: An Upgrade on Traditional Machine Learning for Medical Classification
Ratnaparkhi Gaurang Vinod, Manemoni Arun Kumar, Chinthalapudi Vaahnitha Chowdary, K Suvarchala · 2025
Pancreatic cancer is one of the most fatal cancers, and medical doctors basically face major problems in its early diagnosis. This article throws light on the work involving the application of Kolmogorov-Arnold Networks (KANs), where we deploy a new machine learning paradigm in the diagnosis of pancreatic cancer. The process of classification basically happens in pancreatic cancer using urinary biomarkers. Unlike Logistic Regression and XGBoost, models have struggled with complex and non-linear medical data, KANs can manage large and complex datasets with ease. Using a dataset of urinary biomarkers, trained a KAN model with EBM techniques for feature selection to increase the model's interpretability and feature relevance. The model presented in this paper showed superior performance compared to traditional models in terms of high accuracy and generalization capability. KANs therefore will be very essential in handling efficiently the challenge presented by the vastness of data in medical studies regarding pancreatic cancer detection. Findings from the studies indicate a better potential in reliable use for progression in developing improved detection of pancreatic cancer using the KAN method. The publication contributes to current research that makes innovative uses in machine learning a reality for improvements in diagnosis as well.