Deep learning for crop yield prediction: a systematic literature review

Deep Learning has been applied for the crop yield prediction problem, however, there is a lack of systematic analysis of the studies. Therefore, this study aims to provide an overview of the state-of-the-art application of Deep Learning in crop yield prediction. We performed a Systematic Literature Review (SLR) to identify and analyze the most relevant papers. We retrieved 456 relevant studies of which we selected 44 primary studies for further analysis after applying selection and quality assessment criteria to the relevant studies. A thorough analysis and synthesis of the primary studies were performed with respect to the key motivations, the target crops, the algorithms applied, the features used, and the data sources used. We observed that Convolutional Neural Network (CNN) is the most common algorithm and it has the best performance in terms of Root Mean Square Error (RMSE). One of the most important challenges is the lack of a large training dataset and thus, the risk of overfitting and as a result, lower model performance in practice. For researchers in this field, it is valuable to indicate the current challenges and the possibility for further research, because they tend to focus on the importance of missing research topics.

Automated summary

This study provides an overview of the state-of-the-art application of Deep Learning in crop yield prediction through a systematic literature review. The results show that Convolutional Neural Network (CNN) is the most commonly used algorithm and has the best performance. The lack of a large training dataset poses a significant challenge, leading to the risk of overfitting and lower model performance.