Alzheimer's is a type of dementia caused by aging that begins with mild memory loss and progresses to total loss of mental and physical functioning. Risk awareness allows patients to take preventive action even before irreversible brain damage occurs, an accurate diagnosis of Alzheimer's Disease (AD) is crucial to patient care, particularly in the early stages of the disease. The majority of machine detection techniques are limited by congenital observations, despite the fact that numerous recent researches have employed computers to diagnose AD. However, it requires a lot of time and the result is inaccurate. In the proposed system for AD identification, deep learning techniques address these challenges. Initially, AD MRI are collected and then pre-processed using nearest average interpolation, BPDHE, and WRA-Net, which are employed to improve the quality of the data. Subsequently, a hybrid CapsNet-DNN model is employed for feature extraction and predicting the disease from the pre-processed data. Utilizing performance metrics like recall, accuracy, precision, and specificity, the efficacy of the proposed method is assessed; the results show 91%, 98%, 96%, and 98%. This shows that the suggested model operates consistently and accurately predicts Alzheimer's disease.