The neurodegenerative disorder Alzheimer’s disease (AD) is the leading cause of dementia worldwide. There is currently no known method of halting or slowing the development of Alzheimer’s disease. There is growing consensus that preclinical and moderate cognitive impairment (MCI) stages of the disease are when disease-modifying medicines should be most effective. Different but complimentary information can be gleaned from multi-modality imaging techniques including MRI, FDG-PET, amyloid-PET, and the recently released tau-PET, making it possible to diagnose AD and provide a prognosis (chance of turning to AD) at these early stages. This research study has implemented a modified dragonfly algorithm based on deep learning concept to predict Alzheimer’s disease. The model is broken down into its first two steps—training and testing that are offered to individuals, followed by the collection of data from those individuals and the use of the Welch technique for the model’s preliminary processing. Initially, the collected dataset will be given to the model, after pre-processing the dataset, the improved dragonfly algorithm is used to perform feature selection and classification based on Convolution Neural Network (CNN). The proposed algorithm is implemented in python. The results of the experimental research indicate that the proposed model has attained a maximum aggregate of about 95.65%, sensitivity 94.19%, specificity 95.06%, and F-score 95.29% The application of the proposed technique with a supervised machine learning strategy has made the proposed model to achieve high accuracy.