DIAGNOSIS OF LUNG CANCER USING ENSEMBLE LEARNING APPROACH

ABSTRACT

Lung cancer is one of the deadliest diseases caused by cancer. The treatment and diagnosis of these disease is still a herculean task with regards to prediction. This is due to the scarcity of medical practitioners and the lack of medical in those areas. Over the years, experts both in medical and computing have been engaged in the use of computational intelligence (machine learning algorithms) for the prediction of lung cancer as second opinion for decision making. However, there is need for the improvement of the use of machine learning techniques as there’s no onesize-fit all model that solves all problems. In this study, feature selection methods and ensemble methods are adopted in the development of a prediction model for lung cancer. The filter selection techniques are chi-square, mutual information and correlation methods and the embedded feature method used in the recursive feature elimination support vector machines(RFE-SVM) the support vector machines, decision trees, k-nearest neighbor and naïve bayes were used the lease classifiers for bagging, boosting and stacking method. In this project, the adequacy of using an ensemble of filter-embedded approach and ensemble of classifiers rather than a single filter was demonstrated on the lung cancer dataset with better classification accuracies. the bagged SVM, Decision Trees (DT) and K- Nearest Neighbors (KNN), boosted decision tree and Gaussian naïve bayes, stacking and random forest achieved 99.9%, 95.7%, 99.2%, 100% ,94.8%, 100% and 98.6%. using the ensemble hard voting gave 100% in this study