Evaluation of Unsupervised Learning Modeling with Parallel Processes

In an attempt to optimize computation resources for unsupervised learning techniques, including k-means and k- means++, different methods of parallelization are proposed depending on the scale of business and demands. A set of three different methods (multiprocessing and OpenMP based mechanisms) are used to conduct parallel processes of computing k-means and k-means++ while using six different datasets with varied sizes on two distinct environments, namely Azure cloud- based instance, and a local machine computer. The results of this experiment show that there is no clear decreasing trend of computation running time while processing k-means and k- means++ in different approaches of parallelism.