Deep Learning-Driven Disease Prediction System in Cloud Environments using a Big Data Approach

Abstract

Disease detection has gained great significance in the big data cloud systems as health organizations try to accrue manifold benefits from massive amounts of data to achieve better outcomes in patients. The ability to detect diseases with better efficiency can achieve timely interventions, better resource allocations, and enhanced treatment strategies. However, there are several challenges with the implementation of such systems regarding the integration of different kinds of data, handling imbalanced datasets, and ensuring that real-time processing is capable of supporting clinical decisions. This paper addresses these challenges by proposing a novel disease prediction model on a cloud environment, incorporating the strengths of both Recurrent Neural Networks and Convolutional Neural Networks. The proposed RNN+CNN-based model can effectively capture temporal dependencies and spatial features from different data sources, like medical images and time-series patient data. This model has given improved results through extensive experimentation with different key evaluation metrics like accuracy, precision, recall, F1-score, and AUC-ROC, which are quite higher than other disease prediction systems. Besides, the proposed model is optimized for light speed in inference times, reaching an incredible processing speed of 0.4 seconds per prediction. Thus, the Big Data approach provided within the cloud infrastructure will ensure the predictability and accessibility without bounded local computational resources, opening advanced predictive analytics capabilities to healthcare practitioners.