Optimized-Deep Neural Network (Op-DNN) Regression Technique for Lung Cancer Survivability Prediction

Abstract

Lung cancer is the foremost cause of cancer deaths globally. If lung cancer is successfully identified and predicted in its early stages can reduce many treatment options and also increase the survival rate of the patient. In this study, the Electronic Health Record (EHR) data were examined, and the survivability ratewas predicted for lungcanceraffected patients. If the patients survived for more than one year, the status was predicted (Dead or Alive) by applying classification model. Considering the patients who have survived, a novel Optimized-Deep Neural Network (Op-DNN) regression modelis implemented to predict the number of months that the patient would survive. This will help the doctor for providing chemotherapy to those patients who survived at a specific period of time. Optimal-DNN (Op-DNN) regression model was developed and the evaluation and performance of this model is comparedwith Artificial Neural Networks (ANNs) regression mode. The models were varied by increasing and decreasing the hidden layers and the number of nodes in the hidden layers. Metric evaluation is done by using the activation function Rectified Linear Unit (ReLU),Adaptive Moment Estimation (ADAM), Mean Squared Error (MSE), Mean Absolute Error (MAE) and higher R-squared score(r²)The fine-tuned model having an optimal balance of bias and variance was used to predict the survivability of lung cancer patients( in months).