期刊名称：International Journal of Computer and Information Technology
印刷版ISSN：2279-0764
出版年度：2018
卷号：7
期号：4
页码：170-175
出版社：International Journal of Computer and Information Technology
摘要：The classification is one of the important directions of
supervised learning. The prediction of highly sparse data is a
challenge and an open issue. In this regard, this study conducts
comparative analysis of eight machine-learning algorithms for
classifying healthcare data (i.e., Heart Diseases). The eight
classifiers used in this study are: 1) Naïve Bayes (NB), 2) Single
Conjunctive Rule Learner (SCRL), 3) Radial Bias Function
(RBF), 4) Decision Tree (DT), 5) K-Nearest Neighbor (k-NN), 6)
Multilayer Perceptron (MLP), 7) Random Forest (RF), and 8)
Support Vector Machine (SVM). In order to obtain better
classification outcomes, ensemble-learning methods such as
bagging, boosting, decorate, voting, random sub space and
dagging have also been used in conjunction with considered
classification models. The experimental results have been
validated using 10-fold cross validation method. It has been
revealed in results that SVM performed better in both cases: i)
Simple classification model, and ii) Classification model with
ensemble-learning methods. The accuracy of SVM, in both cases,
achieved 86.13% being the top classifier among the considered
models. The RBF produced second higher accuracy 83.82% and
third MLP as 83.5%. The study indicates that the classification
models in conjunction with ensemble-learning methods can
significantly enhance the predictive outcomes and scalability of
classification schemes, which is of practical importance when
used for healthcare data.
