Systematic study on crop yield prediction methods using deep learning

Abstract

DNNs (Deep Neural Networks) have estimated agricultural but lack comprehensive analysis of findings. The article gives an overview of the existing literature available in DNNs in predicting agricultural productions. This work’s SLRs (Systematic Literature Reviews) were executed to assess most relevant studies. The searches resulted in 456 relevant studies based on quality assessments of which 44 primary studies were selected for this analysis. This work’s examinations include data sources, key motives, targeted crops, algorithms used and features selected. Predominant usage of CNNs (Convolution Neural Networks) was found in the studies as their performances in terms of RMSEs (Root Mean Square Errors) are the best. One serious issue discovered was the absence of large training datasets which give rise to over fits of data and poor model performances. Since, researches look for gaps in studies; it is beneficial to highlight present issues and potential areas for further researches.