Sensitivity Analysis of a Multilayer Perceptron Network for Cervical Cancer Risk Classification

Abstract

In recent times, deep learning and machine learning algorithms are being applied to aid in the diagnosis of cervical cancer to facilitate early diagnosis and reduce mortality rates. This paper concerns sensitivity analysis on an existing cervical cancer risk classification algorithm with respect to the number of epochs, number of neurons in the input layer (NIN), and number of neurons in the hidden layer (NNIHL). Sensitivity analysis is used to analyse the performance of the cervical cancer classification algorithm based on a Multilayer Perceptron Network when changes are made to the setup or architecture of the algorithm. Experimental results reveal that that the algorithm yields a high accuracy when it is trained at 500 epochs with eight input neurons and 100 or 300 or 500 neurons in the hidden layer. We also analyse the execution time of the algorithm under the varied parameters and discover that higher values for NIN, NNIHL and number of epochs all yield longer execution times. These results can aid in the successful application of deep learning for cervical cancer risk prediction.