Evaluation of Machine Learning and Ensemble Methods for Complications Related to Port a Cath

Abstract

This study examines and compares the accuracy of several machine learning algorithms (ML) for predicting Port a Cath complications using ensemble learning (EL) approaches, in order to aid doctors in selecting the most effective treatments for saving lives. The data collection of 794 instances served as the primary database for the training and testing of the built system. 10-Fold Cross-Validation has been applied to expand the data set, which would not have been possible otherwise. The techniques are developed using the Python language. Different classifiers, namely Decision Tree (DT), K-Nearest Neighbor (K-NN), Naïve Bayes (NB), Multilayer Perceptron (MLP), and Stochastic gradient descent (SGD) have been employed. The dataset has also been used for the ensemble prediction of classifiers, bagging, voting, and stacking. The study’s findings indicate that using the voting strategy in conjunction with the MLP, NB, KNN, DT, and SGD methods yields results with an overall accuracy of 92.5% higher than those obtained with the other methods described above.