Performance Analysis of Machine Learning Algorithm on Cloud Platforms: AWS vs Azure vs GCP

Abstract

The significance of adopting cloud technology in enterprises is accelerating and becoming ubiquitous in business and industry. Due to migrating the on-premises servers and services into cloud, companies can leverage several advantages such as cost optimization, high performance, and flexible system maintenance, to name a few. As the data volume, variety, veracity, and velocity are rising tremendously, adopting machine learning (ML) solutions in the cloud platform bring benefits from ML model building through model evaluation more efficiently and accurately. This study will provide a comparative performance analysis of the three big cloud vendors: Amazon Web Service (AWS), Microsoft Azure and Google Cloud Platform (GCP) by building regression models in each of the platforms. For validation purposes, i.e., training and testing the models, five different standard datasets from the UCI machine learning repository have been employed. This work utilizes the ML services of AWS Sage maker, Azure ML Studio and Google Big Query for conducting the experiments. Model evaluation criteria here include measuring R-squared values for each platform, calculating the error metrics (Mean Squared Error, Mean Absolute Error, Root Mean Squared Error etc.) and comparing the results to determine the best performing cloud provider in terms of ML service. The study concludes with presenting a comparative taxonomy of regression models across the three platforms.