Towards Data Science in Agriculture with Big Data Management

Abstract The agriculture data is multivariate, impulsive, and complex; this data must be analyzed and interpreted accurately to address the challenges related to agro-advisory systems, recommendation systems, smart agriculture, information systems, etc. The complex characteristics of agriculture data have made data management more crucial and challenging. Big data management is a very important task to perform to develop data science applications. Data science in the agriculture domain aids in disseminating valuable information to the end users in consumable formats. The goal of big data management is to offer batch, interactive, and iterative computations on a wide range of data problems with low latency and high efficiency. Implementation of data repositories can be a transformative solution not only in terms of data management but also in terms of the business process. The paper proposes an architecture for developing a big data repository for the agriculture domain as a rudimentary step in developing data science applications. The architecture is built and implemented using two levels of data abstraction. As a proof of concept, data are depicted using the web-based REST interface and dashboard. Data science applications such as weather, land usage, market price analytics, and crop price prediction are developed using collected data, as a proof of concept.