PENINGKATAN NEURAL NETWORK DENGAN FEATURE SELECTION UNTUK PREDIKSI KANKER PAYUDARA

Abstract

Breast cancer is increasing in everycountries in the world, especially in developing countries like
Indonesia. Neural network is able to solve problems with the accuracy of data and not linear. Neural
network optimization tested weeks to produce the best accuracy value, applying neural network with
feature selection methods such as Wrapper with Backward Elimination to raise the accuracy produced by
Neural Network. Experiments conducted to obtain optimal architecture and to increase the value of
accuracy. Results of the research is a confusion matrix to prove the accuracy of Neural network before
optimized by Backward Elimination was 96.42% and 96.71% after becoming optimized. This proves the
estimation of feature selection trials using neural network-based method Backward Elimination more
accurate than the individual neural network method.