Comparative Analysis of Adam and RMSprop Optimizer on Artificial Neural Network for Clinical Data-Based Classification of Lung Cancer

Abstract

Lung cancer is one of the leading causes of global death that demands early and precise detection. Advances in Artificial Intelligence technology, especially Artificial Neural Networks (ANNs), can support the disease classification process through medical data. This study focused on evaluating the performance of the ANN model in classifying lung cancer as well as comparing the impact of the Adam and RMSprop optimizers on model performance. The data used was in the form of clinical records of lung cancer patients in CSV format as many as 310 entries. The research steps include pre-processing of data, normalization with MinMaxScaler, division of training and test data with an 80:20 ratio, construction of ANN models using TensorFlow and Keras, as well as assessment through confusion matrix, accuracy, precision, recall, F1-score, and ROC-AUC. The ANN structure includes an input layer, two hidden layers with a ReLU activation function, and an output layer with sigmoids for binary classification. The findings showed that Adam's optimizer delivered the best results with 97 percent accuracy, 98 percent accuracy, 98 percent recall, and nearly 1 AUC. While the RMSprop optimizer produces 95% accuracy. The findings confirm that optimizer selection affects the performance of ANN classification on lung cancer data. It is hoped that this research can be a reference in the development of a medical decision system based on Artificial Intelligence