Adaptive Deep Belief Neural Networks for Pre-Term Birth Clinical Record to Sense Neonatal Apnea Level Classification

Abstract

A Deep learning method has been presented to identify the risk factors for Pre-term Birth (PTB). Premature birth is one of the most important factors that affect the death of the infant. The existing method analyzes the Very low birth weight and pre-term infants more than 1500 grams is a high risk of developing intraventricular bleeding, which is a major cause of brain damage in premature infants. The previous method shows time complexity, and feature selection is being provided the highest error rate taken. To overcome the issues in this work proposed the method, Adaptive Deep Belief Neural Networks (ADBNNs) algorithm analysis to using the Softmax Late-Onset Sepsis (SLOS) function for utilizing the risk factors. Initially, the Pre-processing for non-redundant data from data begins to function using the Dynamic Ensemble Selection (DES) algorithm, which reduces the relevant values of the dataset. The proposed method Adaptive Deep Belief Neural Networks (ADBNNs) algorithm, was used to classify results based on the feature extracting information contained in the original set of features. The classification results show the Neonatal Apnea Level Classification should be calculated and combined with the Risk factors analysis based on the Softmax activation function classified the hidden layer function called Autoencoders Deep Belief Network. Hidden layers or invisible layers are not connected and are conditionally independent. Experimental results show that to perform a defect classification with the proposed method, an ADBNNs would isolate the optimal features of the individual with minimal network training time, and ultimately, the individual in the prediction and reducing the error rate, time complexity, and time complexity improving the accuracy.