Klasifikasi Kanker Tumor Payudara Menggunakan Arsitektur Inception-V3 Dan Algoritma Machine Learning

Abstract

Breast cancer is a disease that arises due to breast tissue cells that grow abnormally and continuously. This disease is a disease with a large increase in number of around 13 million per year, with a mortality rate of 9.6% from a total of 65,858 cases. Early detection of breast cancer for prevention needs to be done, with the hope that breast cancer is easier to treat and cure and can even be prevented before it enters an advanced stage. In this research, build a model with transfer learning technique for breast cancer classification. There are 4 methods tested, namely Inception-V3 feature extraction with the Radial Basic Function Neural Network classification method, FeedForward Neural Network, Logistic Regression and feature extraction by making changes to the hyperparameter layer. This study compares the four models to get the best one to solve the problem of breast cancer classification. The data used in this study are breast cancer image data with a zoom scale of 40X, 100X, 200X and 400X. The dataset was sourced from The Laboratory University of Parana with P&D Laboratory Pathological Anatomy and Cytopathology, Parana, Brazil. The results of this study indicate that the Inception-V3 feature extraction method with the Logistic Regression classification method on the 40X zoom scale data provides the best accuracy (93.00%), precision (94.00%), and recall (91.00%) F1-score (92.00%).