Performance Evaluation of Machine Learning Models for Multi-class Lung Cancer Detection

The main reason for the increasing number of deaths worldwide is cancer, among them the main cause of highest mortality rates is lung cancer. Approximately 85% of the male and 75% of the female suffer from lung cancer. The cancer cells will keep on growing and multiplying leading to the development of tumor. The rapid growth of these cells can spread to other parts of the body, this is known as Metastases. Recognition of cancer at its final stage has barely any chances of getting complete treatment. It might lead to the death of the patient. Consequently, early recognition of the cancer before its final stage is highly crucial to increase the survival rate of the patient. For early detection, several machine learning techniques are used to clear the way for fast treatment of the disease. The dataset consists of different attributes such as smoking, alcohol consumption, chest pain, shortness of breath etc. Decision tree, logistic regression, SVM, Naïve Bayes, KNN and random forest are the various ML classifiers applied to the dataset. The classification models are analyzed for different test and train ratios and the obtained accuracy, precision, recall, error rate, specificity, F-measure and testing time are noted. This process is carried for both binary and multi class classification. Multiclass is considered as three class classifications, i.e. high, low and medium execution time. Therefore, the SVM classifier using Machine Learning technique can be applied to detect the presence of the disease. Hence, it helps the doctors in identifying it. By doing so, early diagnosis can be performed and required precautions can be taken.