Evaluating the Accuracy of Patient Length of Stay Predictions Through a Comparison of Deep Belief Networks and Convolutional Neural Networks

This study aims to predict patient length of stay in healthcare centers, analyze discharge patterns, and forecast unplanned readmissions to the Intensive Care Unit using machine learning algorithms. Specifically, it compares the effectiveness of SVM and Logistic Regression (LR) algorithms, and evaluates the performance of Novel Deep Belief Networks (DBN) and Convolutional Neural Networks (CNN). The research involved analyzing 47 samples with two distinct groups of algorithms. A G-power of 80%, a significance threshold of 0.05, and a 95% confidence interval were employed to ensure accurate results. Patient health data were collected from various recent web sources. The results indicated that the Novel Deep Belief Network algorithm achieved an accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 0. 2 5 \%}$</tex> in predicting patient length of stay and discharge, while the Convolutional Neural Network algorithm achieved an accuracy of 87.11%. Although CNN performed effectively, the difference in accuracy between DBN and CNN was not statistically significant <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(p=0.248; p&gt;0.05)$</tex>.