Despite the high incidence of Alzheimer’s disease (AD), there is no cure for AD yet. Therefore, early identification of individuals at higher risk of developing AD becomes critical, as this may provide a window of opportunity to adopt lifestyle changes to prevent or delay the onset of the disease. We propose a novel approach to developing prediction models using Feed forward Deep Neural Networks. Our models are built using the EPAD LCS v.IMI dataset. We extract a combination of brain imaging, genetics, cognitive and lifestyle features from the dataset to build the prediction models. The prediction is based on the ATN classification framework, with prediction categories of Healthy, Suspected Non-Alzheimer’s Pathology (SNAP), and Dementia due to AD continuum. We built a total of 6 prediction models, of which 4 are based on classic Machine Learning (ML) and 2 are Deep Learning (DP) approaches. The best DP model outperforms the classic ML model by F1 score of 14% and AUC score of 13%. We have demonstrated that our Deep Learning based model has the potential to be deployed as a screening model to predict dementia risk at early stage of the disease.