Diagnostic Accuracy of Machine Learning Algorithms for Hepatitis A Antibody

Abstract

Purpose: The objective of this study was to develop a model for predicting the positivity of hepatitis A antibody based on nationwide health information using a machine learning technique. Methods: We used a data set that included the records of 4,626 samples. the data was randomly divided into a training set 80% (3,701) and validation set 20% (925). Customized sequential convolutional neural network (CNN) model was used to predict the positivity of hepatitis A antibody. The loss and accuracy of this model was calculated. Results: This model has 12-input and 2-concatenate and 3-dense layers. The total parameters of this model were 1,779. The accuracy quickly reached to over 85% validation accuracy in 50 epochs. The train loss, train accuracy, validation loss and validation accuracy of this model were 25.4%, 89.5%, 29.0%, and 87.2%, respectively. Conclusion: The model derived from the sequential CNN model exhibited a high level of accuracy. This model is a useful tool for predicting the positivity of hepatitis A antibody.

Keywords

• hepatitis a virus
• antibodies
• machine learning
• algorithms