Ensemble Deep Learning based Lung Cancer Classification Model using Gene Expression Data

Background: Globally, lung cancer is the deadliest form of the disease. Genetic variability is one of the elements that influence an individual's vulnerability to lung cancer, according to epidemiological research. Asian women, smokers or not, have a higher risk of acquiring cancer because of genetic abnormalities, according to a recent investigation from the US National Cancer Institute that involved 14,000 Asian women. A superior approach for classifying lung cancer was presented in recent studies to address the aforementioned issue. In this study, the data scale is first normalized utilizing min max normalization, which is accomplished by data pre-processing. Methods: Gene selection is carried through employing Improved Whale Optimization Algorithm (IWOA). An Enhanced Convolutional Neural Network (ECNN) is employed for lung cancer categorization. However, lung cancer classification using single algorithm produces insufficient accuracy. This required the need for development of ensemble models. To evade this issue,input data scales are normalized based on Z score normalization model. Once the normalization is done, significant genes are selected from these normalized gene samples using Modified Chicken Swarm Optimization (MCSO). Results: Finally, ensemble of ECNN, VGG16 and ResNet50 models are employed for lung cancer classification. Ensemble learning is performed in this work using majority voting. Conclusion: The suggested approach outperforms various alternatives in the field of accuracy, according to the findings.