Machine Learning Lifecycle for Earth Science Application: A Practical Insight into Production Deployment

Enterprises are making machine learning for production as an integral part of their future roadmaps and Earth science domain is no exception. However, there is common problem in transitioning machine learning from science to production due to a major difference in constructing a model versus deploying it for people to use to make decisions. Phases of machine learning lifecycle that includes model transition to production using a successful application is discussed.

1. INTRODUCTION

Machine learning (ML) has enabled data-driven solutions to numerous science problems. Earth science domain presents unique sets of problems that are increasingly being solved using data driven approaches. These problems are diverse in nature and the subsequent analysis of Earth science data typically differ from standard problems encountered in other domains. ML for Earth science is challenging for many reasons namely the spatio-temporal and multivariate nature of the data. However, Earth science data is mostly open and publicly available via catalogs unlike other domains hindered by various privacy and copyright concerns. The availability of big Earth science data offers immense potential for ML as evident from numerous research publications lately. However, many of these publications are not ending up as production applications mainly because the data scientists who develop the ML models are now expected to complete the ML lifecycle by deploying and scaling the models in production.