Alzheimer's disease (AD) is a major cause of cognitive decline on a global scale, exerting a substantial impact on healthcare systems and family dynamics. Early and accurate diagnosis is imperative for enhancing patient outcomes and treatment planning. This study employs machine learning (ML) and deep learning (DL) algorithms to address this challenge using a comprehensive dataset that includes demographic, clinical, and behavioral features. The methodological framework encompasses four stages: dataset preprocessing (label encoding, one-hot encoding, SMOTE, and standardization); model implementation (SVM, LR, NB, KNN, DNN, MLP, SGD, Random Forest, Decision Tree, XGBoost, LightGBM, and CatBoost); hyperparameter tuning (grid search cross-validation (GridSearchCV) and Optuna); and model evaluation using metrics such as accuracy, precision, recall, F1-score, and receiver operating characteristic area under the curve (ROC-AUC). The findings indicated that LightGBM exhibited the optimal performance, attaining an accuracy of 96.04% and an ROC-AUC of 0.98, thereby surpassing the other models in both precision and robustness. Feature importance analysis identified MMSE and ADL scores as the primary predictors of Alzheimer's disease, thereby highlighting the dataset's ability to reflect relevant disease attributes. This study underscores the potential of ensemble learning and hyperparameter optimization in enhancing Alzheimer's diagnosis, paving the way for future research endeavors aimed at refining early detection methodologies.