SparkReact: A Novel and User-friendly Graphical Interface for the Apache Spark MLlib Library
Aristeidis Karras, Christos Karras, Agorakis Bompotas, Panagiotis Bouras, Leonidas Theodorakopoulos, Spyros Sioutas · 2022
Visualization is a critical component across every software as it enables users to familiarize themselves with the environment and perform certain tasks with ease. Therefore, straightforward yet interactive and easy-to-understand tools let users’ complex demands be satisfied within minutes. The objective of this work is to give an optimized graphical user interface for the Apache Spark MLlib library to apply machine learning algorithms quickly, conveniently, and effectively. We introduce SparkReact, a responsive graphical user interface that allows users to apply clustering, classification, and regression techniques within just a few mouse clicks by implementing and evaluating a certain algorithm and pre-building the code ready for import to Spark. To evaluate the usefulness of our tool we performed crowdsourcing to two categories, computer experts and ordinary users. The results indicate that both populations were satisfied with the tool at a surprising 98 percent. As per the time required to construct and evaluate a machine learning model, it took approximately 4 minutes using SparkReact while with ordinary methods it took almost 4 times longer. Ultimately, future extensions will seek to provide more algorithmic choices.