Diagnosing Alsheimer´s disease (AD) remains challenging due to its complexity and multifactorial nature. In 2016, a major revision of AD definitions emphasized integrating clinical, neuroimaging, and data analysis to improve diagsostic accuracy. Traditional statistical models and machine learning (ML) techniques have been applied, but treatment strategies remain insufficient when relying solely on clinical phenotypes. Combining clinical features with other data through supervised learning offers a path forward. This study uses the GSE84422 microarray dataset (2006 samples) to identify essential genetic features for AD diagnosis through machine learning models, including Artificial Neural Networks (ANN) and Support Vector Machines (SVM). Systematic feature selection ranked critical genes. ANN achieved superior accuracy with 105 genes, compared to SVM´s 249. An overlap of 86 genes between ANN and SVM indicates model consistency. Key findings revealed discrepancies with prior studies. For example, RIMS3, identified as the fourth most important gene in our analysis, was absent from NETTAG´s prioritized AD-related genes. This divergence highligths how ML models can reveal genetic information overlooked by traditional frameworks. Integrating ML with other data analysis methods could enhance disease diagnosis and treatment strategies, improving the precision of AD detection and management.