Integrative Analysis of Heterogeneous Cancer Data Using Autoencoder Neural Networks

Abstract

Early detection of cancer needs improvement to make real progress against the illness. Machine learning methods especially autoencoder neural networks show promising results in detecting diagnostic patterns from difficult data. Current cancer detection machine learning models handle just one type of medical information which prevents them from seeing biological connections between different types of data. This paper presents an innovative autoencoder neural network method that combines different types of healthcare data for precise cancer treatment. I develop a procedure to create autoencoder neural networks and test network performance on stomach adenocarcinoma cancer patient genetic data. Our approach shows that autoencoder neural networks effectively process biological information while showing how combining multiple data types helps identify cancer conditions.