Prediction of Alzheimer's Diagnosis with Machine Learning and Innovative Feature Engineering

Alzheimer's disease is a leading cause of dementia, presenting significant challenges to healthcare systems globally. Early diagnosis is essential for effective management and intervention, yet traditional diagnostic methods remain invasive, time-consuming, and costly. This study investigates the application of advanced machine learning models, emphasising the role of feature selection techniques, such as RFE (Recursive Feature Elimination) and hyperparameter optimisation, to enhance the early detection of Alzheimer's disease. Among the evaluated models, CatBoost with RFE achieved the highest performance, with an accuracy of 95.81% and an F1-score of 94.00%, demonstrating its robustness and reliability as a diagnostic tool. Random Forest and XGBoost models also showed strong results, particularly when combined with feature importance and RFE. The findings highlight the significant impact of feature engineering and hyperparameter tuning in improving model performance across key metrics, including accuracy, recall, precision, and F1-score. This research underscores the potential of integrating machine learning techniques into medical diagnostics, offering a non-invasive, cost-effective, and efficient approach to Alzheimer's disease prediction. The insights gained from this study lay the groundwork for future advancements in diagnostic models, aiming to improve early detection strategies and patient outcomes, ultimately contributing to the global effort to mitigate the impact of Alzheimer's disease on individuals and society.