Effective factors in diagnosing the degree of hepatitis C using machine learning

Hepatitis C virus (HCV) is a major public health threat, which can be treated if diagnosed early, but unfortunately, many people with chronic diseases are not diagnosed until the final stages. Machine learning and its techniques can be very helpful in diagnosis. This study examines the factors affecting hepatitis C diagnosis using machine learning.

A total of 27 features were used with a dataset containing 1385 records of patients with different grades of HCV. The dataset was clean and preprocessed to ensure accuracy and consistency. To reduce the dimension of the dataset and determine the effective features three feature selection, Pearson Correlation, ANOVA, and Random Forest, were applied. Among all the algorithms, KNN, random forests, and Deep Neural Networks were selected to be utilized, and then their evaluation metrics, such as Accuracy and Recall. To create prediction models, fifteen features were selected for the mentioned machine learning algorithms.

Performance evaluation of these models based on accuracy showed that Deep Learning with Accuracy = 92.067 had the highest performance. KNN and Random Forest had almost the same performance after Deep Learning. This performance was achieved on dataset containing features that were selected by ANOVA feature selection.

Machine learning has been very effective in solving many challenges in the field of health. This study showed that using data-mining algorithms also can be useful for HCV diagnosing. The proposed model in this study can help physicians diagnose the degree of HCV at an affordable and with high accuracy.