The best architecture selection with deep neural network (DNN) method for breast cancer classification using MicroRNA data

Abstract

Abstract Breast cancer is one of the most common causes of death in the world. One way that can be done to reduce the number of death cases is to do early detection using MicroRNA data. MicroRNA is one of the cancer biomarkers that can help in the classification process. MicroRNA can be used to identify whether a cell is a cancer cell or not even in the earliest stages. The deep Neural Network (DNN) method consists of two or more layers of self-learning units (hidden units). The weight of hidden units that are fully connected between two layers can be learned automatically. However, DNN still has a weakness which is the changes in the distribution of each layer’s inputs that cause problems, because the layers need to continue to adapt to the new distribution and produce less optimal accuracy values. This research was conducted to get the best architectural selection of the DNN method used for breast cancer classification using MicroRNA data. The results of the DNN method with the best architecture obtained was 3 hidden layers with 200 hidden units each, 30% dropout rate in the combination of the ReLU, ReLU, ReLU activation functions and the learning rate of 0.04 resulting in the highest accuracy of 94.58% with a specificity of 96.45% and sensitivity of 91.02%.