Prediction of Crop Yield Using Big Data

Abstract

Quantifying the yield is essential to optimize policies to ensure food security. This paper aims at providing a new method to predict the crop yield based on big-data analysis technology, which differs with traditional methods in the structure of handling data and in the means of modeling. Firstly, the method can make full use of the existing massive agriculture relevant datasets and can be still utilized with the volume of data growing rapidly, due to big-data friendly processing structure. Secondly, the "nearest neighbors" modeling, which employs results gained from the former data processing structure, provides a well-balanced result on the account of accuracy and prediction time in advance. Numerical examples on actual crop dataset in China from 1995-2014 have showed a better performance and an improved prediction accuracy of the proposed method compared with traditional ones.