Alzheimer's disease is one of the most frequent neurodegenerative diseases, leading to a disruption in the cognitive process of the human brain. Using this type of dataset to implement machine learning and deep learning techniques is a common approach for detecting and classifying Alzheimer's disease. In this study, we addressed primary research problems, including early diagnosis and accurate classification of Alzheimer's disease, effective preprocessing of imaging and non-imaging data, and identifying the most accurate modelling strategy between machine learning and deep learning techniques. We made an effort to present an advanced neuroimaging-based analysis of Alzheimer's disease early detection, implementing various machine learning and deep learning techniques. We collected our dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI). For our work, we extracted attributes such as structural MRI, clinical assessments, and cognitive data from our collected dataset. In this study, we employed machine learning and deep learning techniques separately to evaluate their precision and accuracy in detecting and classifying Alzheimer's detection, which led to identifying the most optimized results. We utilized a custom CNN and Self-Attention (SA) model, along with DenseNet, ResNet-50, and VGG-16, to implement deep learning techniques. For machine learning techniques, we employed Naive Bayes, Support Vector Machine (SVM), K-Nearest Neighbours (KNN), CatBoost, and Gradient Boosting models. To compare our method with state-of-the-art strategies, we used metrics such as accuracy, precision, and F1 scores. Our approach outperformed existing machine learning and deep learning models. In our approach, CNN and Self-Attention model achieved an accuracy of 98.20%.