Parallel and extremely fast learning neural network for healthcare big data applications

The incorporation and growing uptake of machine learning (ML) algorithms in big data in healthcare has transformed real-time medical diagnostics, predictive analytics, and clinical decision-making. The use of machine learning in healthcare big data, however, is fraught with computational difficulties, especially when it comes to effectively processing and training on massive amounts of high-velocity data produced by healthcare institutions throughout the globe. To address these challenges, this study introduces a novel Fast Learning Network (FLN) framework designed to overcome these limitations by introducing an eXtremely Fast Learning Network (XFLN). The suggested method calculates output weights directly, removing the need for slow backpropagation and allowing for quick training on regular computers. This work develops a Tikhonov-regularized fast pseudoinverse algorithm to tackle computational bottlenecks and enhance numerical stability in high-dimensional, ill-conditioned healthcare data. Additionally, the proposed Parallel and eXtremely Fast Learning Neural Network (PXFLN) improves scalability by allowing multiple processes to run at the same time in systems with multiple cores or clusters. Experiments on various healthcare datasets show that PXFLN improves predicted accuracy while achieving speedups of up to 280 times over conventional techniques. Additionally, it has strong scalability, reducing processing costs per sample by as much as 90% as parallelism rises. Thus, the FLN-based architecture provides a scalable, accurate, and effective way to analyze massive amounts of healthcare data in real time. Ultimately, PXFLN supports the world in achieving the Sustainable Development Goals (SDGs) by enhancing processes to make them more efficient and accelerate the acquisition of clinical insights. This is more so the objectives that concentrate on good health, innovation, and infrastructure.