Penerapan Algoritma Backpropagation Untuk Memprediksi Jumlah Pasien Pra-Kanker Serviks (Studi Kasus Di Puskesmas Padang Pasir)

Abstract

According to the Global Cancer Observatory report in 2018, Indonesia ranks second in the largest cervical cancer cases in the world. 80% of cervical cancer sufferers come in an advanced stage, and 94% of patients of these cases die within 2 years. Preventive action from the government is expected to reduce the number of sufferers. By knowing the increase in the number of patients, the government can take action on what must be done to reduce the number of pre-cervical cancer patients. Data on the number of patients with cervical pre-cancer is continuous, so the method needed to make predictions is a complex method and study the uncertainties in each period that can be accommodated with the Artificial Neural Network (ANN) Backpropagation algorithm. Backpropagation architecture used is 5 input layers, 5 hidden layers, and 1 output layer, with a learning rate (lr) of 0.5; constant momentum (mc) 0.3; the Mean Square Error (MSE) value of network training is 0.001 with the logsig activation function for hidden layers and purelin for the output layer. Resulting in a 5-5-1 network architecture pattern with the epoch = 322 processes and the achievement of MSE when testing with MSE = 0.001 with an accuracy of 99.999%.