IMPROVED LUNG CANCER DETECTION USING MACHINE LEARNING TECHNIQUES

Abstract

Lung Cancer have become one of the most common cause of death among human beings. Many human beings die early because of lung cancer. The early detection of lung cancer is tough due to the structure of cancer cells and less awareness among human beings. Diagnosis of lung cancer is done using various tests like imaging tests, sputum cytology and biopsy, which are costly and time taking. Classifying lung cancer is not an easy task and needs experienced physicians and a lot of money. Cancer recurrence in recovered patients leads to a high cost, and not everyone can afford it. We have used lung cancer data from Wu, Jiangpeng et al. [11] as it is an unbalanced dataset. We used various evaluation parameters like Accuracy, Confusion matrix, AUC- ROC [6] and FNR, which gives us more insight. XGBoost provides the best accuracy of 92.16% for lung cancer. We have used various Machine Learning classification techniques under the library of scikit-learn like KNN, Logistic Regression, XGBoost, Gaussian Naive Bayes, Decision Tree and SVM. Different algorithms under the library of scikit learn have been used to select the features, and only those features that are important to our model are selected. Our models achieved more accuracy score and sensitivity than Wu, Jiangpeng et al. [11] for lung cancer. Parameter tuning has helped to improve the performance of the model. In various phases of machine learning pipeline, we have improved the result of Wu, Jiangpeng et al. [11]. Starting from preprocessing to the development of the model.