Classification of Breast Cancer using Stacked Autoencoders and Performance Comparison with Classical Machine Learning Methods

44906

Breast cancer is among the most dangerous cancer types that cause many deaths every year. Early diagnosis states play a constructive role in cancer treatments. Therefore, researchers conduct experimental research on the data that belongs to patients and healty peaople using classification and clustering methods. In addition to machine learning assisted diagnostic studies with the developing technology, there is a significant increase in the use of deep learning methods. In this study, a new model is designed to be used in the classification of breast cancer using stacked autoencoders (SAE) which is a deep learning method. Support vector machines, k-nearest neighborhood, naive bayes, and decision trees methods, which are the most commonly used machine learning methods to compare performance with the designed SAE, were also used in this study. In addition to the accuracy rate metric, the elapsed time (time complexity) during the training and testing stages was calculated in experimental studies. In the experimental studies, the effect of the classification performance was examined by applying the normalization process from the data pre-processing steps. According to the experimental results, the most successful result with 79.31% accuracy rate was obtained by data pre-processing aided SAE. According to the time complexity metric, KNN algorithm is the fastest algorithm in the training process, while the SAE algorithm is the fastest in the test process.

151

Breast Cancer , Deep Learning , Stacked Autoencoders , Data Preprocessing , Machine Learning Algorithms

Erciyes University Journal Of The Institute Of Science an‎d Technology

160

Erciyes University Journal Of The Institute Of Science an‎d Technology