Optimized hybrid machine learning techniques for predictive analysis of lung cancer

The importance of advanced diagnostic tools is underscored by the observation that Lung cancer is still one of the most prevalent causes of cancer-related deathsglobally. Diagnostic accuracy has been shown to be improved with machine learning (ML), and particularly, when incorporating synthetic hybrid features derived from medical imaging data. This research seeks to assess the effectiveness of three standard classification models with synthetic heterogeneous features for lung cancer diagnosis.: Along with SVM and Random Forest, XGBoost was employed. We designed and proposed a distinctive technique for feature extraction to create a strong hybrid feature set to capture local and global attributes in lung images. This approach integrates radiomic, texture-, and shape-based features. This paper presents an approach to early lung cancer detection using machine learning. It addresses the issue of class imbalance using the Synthetic Minority Over-sampling Technique (SMOTE), and improves the performance of the model using Bayesian optimization. This paper evaluates three different classifiers: Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), both with and without hyperparameter optimization. With accuracies of SVM: 95.74%, RF: 96.11%, and XGBoost: 95.92%, the optimized models showed how well SMOTE, machine learning algorithms, and hyperparameter tuning work together to increase prediction accuracy for the diagnosis of lung cancer.