AI-based Lung Cancer Prediction: Development, Comparison, and Evaluation of Machine Learning Models

Artificial intelligence plays an important role in early diagnosis of various pathological disorders. This research has precisely focused on comparing three machine learning models, Decision Tree, k-Nearest Neighbors, and Naïve Bayes, for the prediction of lung cancer. Data has been preprocessed and divided into test and train subsets. Machine learning models are designed from scratch and tested on train and test subsets. Cross-validation and the analysis of train and test accuracies are used to confirm the consistency of results. Furthermore, evaluation parameters vary in their applications and emphasis is to minimize the chance of any false negative predictions, hence recall has been selected as an appropriate assessment tool for this three-class classification problem. K-NN has outperformed with 0.998 train accuracy, 0.995 test accuracy, and 0.990 recall. Moreover, consistent results in cross-validation and only 0.3% difference between the test and train accuracies, have outcasted the chance of overfitting. Lung cancer prediction using machine learning is important because it increases survival rates. As we embark on this study, it helps to define a healthy lifestyle to minimize the risk of lung cancer. In future, this classification model could be deployed in personalized treatment strategies, which could improve patient outcomes.