Federated auto-encoder and XGBoost schemes for multi-omics cancer detection in distributed fog computing paradigm

Abstract

The digital healthcare paradigm has significantly improved based on distributed fog and cloud networks for cancer detection with multiple classes in recent years. The paradigm allows the collecting and training of cancer data on various computing nodes to make optimal cancer detection with their classes. In paradigm, multi-omics approaches (such as RNA, miRNA, and methylation) and machine learning techniques have achieved remarkable results in predicting cancer with different features. However, the digital healthcare paradigm for cancer detection concerning infrastructure still faces challenges related to security, execution delay, and improving the accuracy of cancer prediction in existing studies. These limitations affect the overall cancer detection results with more accuracy. In this paper, we are handling the research mentioned above limitations. We present a new paradigm for cancer detection with more accuracy, less processing delay, and more security based on fog cloud heterogeneous computing nodes. We present the Multi-Cancer Multi-Omics Clinical Dataset Laboratories (MCMOCL) Schemes to predict multi-cancers with multiple classes and consist of federated learning, auto-encoder, and XGBoost methods. The main objective of this study is to improve accuracy, reduce execution delay, and improve the security among heterogeneous cancer clinics in the architecture. Simulation results show that MCMOCL outperformed all existing machine models in terms of accuracy by 98% , processing delay by 61%, and security for the multi-classes types of cancers in heterogeneous fog cloud paradigm.