The process of big data classification has been extensively explored for the past decades, being it essential for Machine Learning. Likewise, performing statistical analysis and investigating the optimization factors of applied algorithms, while evaluating datasets based on recent experience, are important as well in order to specify the best case scenarios parameters that enhance their accuracy. In this paper search on and implementation of classification algorithms, such as the state-of-the-art k-Nearest Neighbors (k-NN), the alternative k-Nearest Neighbors on Feature Projections (k-NNFP) algorithm and Vote Feature Intervals (VFI), over big medical datasets from UCI Machine Learning Repository, is presented. These algorithms, moreover, are statistically analyzed on a standardized Arrhythmia Dataset [1] in order to extract the scenarios that enhance the precision execution for multiple k-fold cross validations and the number of nearest neighbors. Additionally, feature selection and extraction methods are implemented in order to exploit and establish best-case scenarios.