Image Processing-based Performance Evaluation of KNN and SVM Classifiers for Lung Cancer Diagnosis

Abstract

It is important to note that the cure rates in cases of advanced stages of lung cancer are remarkably low, which stresses out the importance for early detection as means to increase survival chances. A strong area of focus when it comes to increased research in the lung cancer diagnosis is the search for ways through which this disease can be identified at its early stages. The methodology described below is proposed as a means to facilitate early detection of lung cancer. There are two phases in this approach. The study deals with effectiveness of three types of classifiers K-Nearest Neighbors (KNN), Random Forest and Support Vector Machine (SVM) to identify cases related to lung cancer via relevant medical data assessment. In this application, the eval axis performs profiling or measures the accuracy of applying these classifiers and discriminating between cancerous instances versus non-cancerous ones within the dataset. To rate the adequacy of classifiers in distinguishing classes, performance metrics like accuracy, precision, recall and F1-score are used. Furthermore, the research compares KNN, Random Forest and SVM, explaining their specific advantages as well as disadvantages logically referring to how they can or cannot be applied while detecting lung cancer. This investigation shows helpful results in suggesting the possibility that machine learning techniques could assist to identify lung cancer as exact and timely as possible, providing more successful diagnostic procedures and patient outcomes. The experimental findings show that SVM gives the best result at 95.06%, KNN comes second with a percentage of 86.89.