Predictive identification of oral cancer using AI and machine learning

Oral cancer remains a significant global health issue, often diagnosed late due to limitations in traditional diagnostic methods. This study explores the application of artificial intelligence (AI) and machine learning (ML) to enhance the early detection and diagnosis of oral cancer. We investigated three data cleaning techniques missing value imputation, outlier detection, and normalization and assessed their impact on model performance. Using convolutional neural networks (CNNs), support vector machines (SVMs), and random forests, we compared the effectiveness of these techniques in improving diagnostic accuracy and mean squared error (MSE). The results demonstrated that normalization, specifically min-max scaling, was the most effective method, leading to the highest accuracy (94 %) and the lowest MSE (0.013) for CNN models. Outlier detection also improved performance, achieving 93 % accuracy and an MSE of 0.014, while missing value imputation resulted in a lower accuracy of 92 % and an MSE of 0.015. These findings underscore the importance of normalization in preprocessing for machine learning models, highlighting its role in achieving superior performance in oral cancer detection. This study underscores the potential of AI-driven methods to revolutionize diagnostic practices, offering more accurate and timely detection of oral cancer.