Author(s): Ferreira, Diana ; Neto, Cristiana ; Lopes, José ; Duarte, Júlio Miguel Marques ; Abelha, António ; Machado, José Manuel
Date: 2022
Persistent ID: https://hdl.handle.net/1822/80679
Origin: RepositóriUM - Universidade do Minho
Subject(s): Classification; Data mining; Predictive models; Primary biliary cholangitis
Primary Biliary Cholangitis, which is thought to be caused by a combination of genetic and environmental factors, is a slow-growing chronic autoimmune disease in which the human body’s immune system attacks healthy cells and tissues and gradually destroys the bile ducts in the liver. A reliable diagnosis of this clinical condition, followed by appropriate intervention measures, can slow the damage to the liver and prevent further complications, especially in the early stages. Hence, the focus of this study is to compare different classification Data Mining techniques, using clinical and demographic data, in an attempt to predict whether or not a Primary Biliary Cholangitis patient will survive. Data from 418 patients with Primary Biliary Cholangitis, following the Mayo Clinic’s research between 1974 and 1984, were used to predict patient survival or non-survival using the Cross Industry Standard Process for Data Mining methodology. Different classification techniques were applied during this process, more specifically, Decision Tree, Random Tree, Random Forest, and Naïve Bayes. The model with the best performance used the Random Forest classifier and Split Validation with a ratio of 0.8, yielding values greater than 93% in all evaluation metrics. With further testing, this model may provide benefits in terms of medical decision support.
