A Comparitive Study of RNN and Restricted BOLTZMANN MACHINES(RBM) in Predicting HIV Test Results Analyzing Using the EHDS Dataset

This study presents a comparative analysis of Recurrent Neural Networks (RNN) and Restricted Boltzmann Machines (RBM) for the early prediction of HIV test outcomes using the EHDS dataset sourced from Kaggle. With a growing demand for accurate and timely diagnosis in healthcare, predictive models leveraging deep learning can significantly enhance disease prevention and management strategies. The dataset comprised a total of 34,336 samples, divided equally into two groups: Group I (17,168 samples) trained using RNN and Group II (17,168 samples) using RBM. RNNs, known for their strength in handling sequential data, were contrasted with RBMs, which specialize in unsupervised learning of binary data structures. Implementation was carried out in Python, while statistical analysis was performed using clincalc.com, with a power of 80%, alpha of 0.05, and beta of 0.2. The models were evaluated based on accuracy, a key metric for medical prediction systems. Results indicated that the RNN model achieved an accuracy of 98.8%, significantly outperforming the RBM model, which achieved 94.8%. The observed difference was statistically significant, with a two-tailed p-value of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.004(p&lt;0.05)$</tex>, confirming the superiority of RNN in this context. This research highlights the potential of RNN-based architectures in developing more reliable diagnostic support tools in healthcare. The findings emphasize that RNNs can serve as an effective predictive framework for early detection of HIV, thereby supporting timely clinical decisionmaking and better patient outcomes.