A Deep Review to Find the Most Accurate Method Among KNN,RF and MLP for Predicting Alzheimer Disease

Debilitating neurodegenerative conditions like Alzheimer’s disease (AD) are linked to great pain and eventually death. For the purpose of successful treatment planning and individualized therapy, it is imperative that Alzheimer’s disease (AD) and its precursor phases, such as moderate cognitive impairment (MCI), be identified as soon as possible. Single nucleotide polymorphisms, or SNPs, and neuroimaging data, such as positron emission tomography (PET) and magnetic resonance imaging (MRI), have been found to work better together to improve the ability to distinguish between AD/MCI and normal control (NC) participants. But missing data from multimodal sources will always exist, and model learning is complicated by the variety of phenotypic and genotypic biomarkers. Furthermore, most current research uses multistage techniques to forecast illness development and focuses only on classifying brain diseases, ignoring the regression problem. Using the most samples possible inside a k-nearest neighbour, random forest and multi layer perceptron (KNN-RF_MLP) architecture, we provide a unique framework to concurrently solve classification and clinical score regression problems by combining multimodal neuroimaging and genetic data. To impute missing features in partial data, we first apply linear interpolation in our technique. We next employ KNN to extract neuroimaging features from SNPs, MRI, and PET data in order to reduce the variation in phenotypic and genotype data.