Alzheimer's Disease, a neurodegenerative condition, offers a unique opportunity for qualitative analysis when identified in its early stages like Mild Cognitive Impairment (MCI). This research focuses on exploring various multimodal data fusion techniques for Alzheimer's disease classification, drawing from reputable sources like IEEE, Springer Link, and ScienceDirect. Utilizing datasets from ADNI and OASIS addresses four central research questions, focusing on filling missing modality, cross-modality, and feature extraction, along with techniques to reduce neural network size along with in-practice feature selection. Strategies for filling missing modalities, including average and linear interpolation, are analysed. The study also investigates neural network design, exploring variations in pooling layers to reduce space size, and examines various deep learning methods, including integrated classification.