Breast Cancer Classification: Comparison of Bayesian Networks, Multilayer Perceptron, and Boosting Method
Intan Nurma Yulita(*), Shofiyyah Nadhiroh
*Faculty of Mathematics and Natural Science, Padjadjaran University
Jalan Raya Bandung Sumedang KM.21, Jawa Barat 45363, Indonesia
Abstract
Cancer is a dangerous disease that should not be underestimated. The early stages of this disease are often asymptomatic. Early detection of cancer is an important examination so that the disease does not develop into a serious and dangerous disease. This study detected the presence of cancer through five predictors. This study classified the diagnosis results based on five indicators namely radius, texture, perimeter, area, and smoothness. By using these five indicators, the detection was carried out through a classification mechanism using the boosting method. The result had obtained an accuracy of 93.67%. The accuracy was higher than other classification methods such as Bayesian Networks and Multilayer Perceptron. Both of them only obtained an accuracy of 89.63%, and 92.79%, respectively. It showed that the ensemble method mechanism of boosting had proven to be more effective in classifying the presence or absence of breast cancer
Keywords: Breast cancer; Classification; Boosting; Bayesian Networks; Multilayer Perceptron
Topic: Computer Science