همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Enhanced lung cancer detection through SMOTE-ENN resampling with optimized machine learning classifiers and LOOCV

نویسندگان :

Alireza Kazempoor Choobari¹ Sadegh Sulaimany²

1- دانشگاه کردستان 2- دانشگاه کردستان

کلمات کلیدی :

Lung cancer،Classification،Machine learning،Preprocessing

چکیده :

Today, the use of machine learning approaches has proven to be effective in various fields of medicine and healthcare for the diagnosis and prediction of diseases, leading to accurate analysis and classification and assisting healthcare and medical professionals. Lung cancer is one of the deadliest cancers globally, and its early diagnosis is extremely important. In this research, we sought to achieve good results by performing various and methodical data preprocessing steps, such as standardization and using SMOTE-ENN for resampling and noise sample removal, utilizing powerful and optimized machine learning classifiers, and employing the most reliable validation method, namely LOOCV (Leave-One-Out Cross-Validation). We evaluated the classifiers using various evaluation metrics and ultimately achieved the best results with the Random Forest model: Accuracy = 99.38% and F1-Score = 98.85%. The output of this research indicates that appropriate preprocessing and data cleaning by ignoring and removing noisy samples, and using an optimized version of machine learning-based classifiers can lead to an improvement in lung cancer diagnosis.