A Novel Machine Learning Technique to Detect Lung Cancer in CT Images using Auto Color Correlogram Features and Multiple Machine Learning Classifiers

Abstract

 This study presents an analysis of the performance that different machine learning algorithms, with SCHF as the feature extraction method, yield in detecting lung cancer. Categorized models were Naive Bayes Multinomial, Logistic Regression, Additive Regression, Linear Regression, Attribute Selected Classifier, and moreover, Naive Bayes The measured performance features were overall accuracy, precision, recall; F-measure, Cohen’s Kappa, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Relative Absolute Error (RAE), and finally, Root Relative Squared Error (RRSE). Hence, the analysis reveals that Naive Bayes Multinomial model had the highest accuracy (90.62 %) and these performances were significantly better in precision (0.91), recall (0.91) and F-measure (0.78) and lowest error rate among most of the models. The significance of these results is that the recommended SCHF framework is exceptionally suitable for feature extraction while Naive Bayes Multinomial gives the most accurate lung cancer classification. This research highlights how machine learning, particularly higher order approaches, can enhance early detection and, outcomes classification of Lung Cancer to aid clinical management.