Ensemble-based Gene Selection and an Enhanced Deep Multi-Layer Perceptron-based Classification Model for Classifying Alzheimer's disease

Memory and cognitive disabilities have been known to humankind for a long time. However, the pathology and symptoms associated with Alzheimer's Disease (AD) are documented during the current century's first decade. Over the years, AD became the common form of Dementia that has complex pathology and is termed a heterogeneous disorder. The advancements in microarray technology made capturing hundreds and thousands of gene sequences possible. However, the data generated is complex and beyond the understanding of the human brain. AD symptoms are slow but fatal, and the diagnosis is made only through autopsy. Thus, early and accurate diagnosis of the disease is critical. The significant difficulty in handling the gene expression data is the curse of dimensionality or the High Dimensional Low Sample Size (HDLSS) issue. The HDLSS issue demands interdisciplinary research, such as Artificial intelligence, machine learning, etc. This study proposed an ensemble-based feature selection to isolate the necessary genes responsible for causing AD. After selecting the relevant features, the Deep Multi-layer Perceptron (DMLP) is used to classify the AD and non-AD patients. The results are compared with other state-of-the-art feature selection techniques and classification algorithms.