Model-Driven And Data-Driven Approaches For Crop Yield Prediction: Analysis And Comparison
Lists
Only you see this
Abstract
Crop yield prediction is a paramount issue in<br> agriculture. The main idea of this paper is to find out efficient<br> way to predict the yield of corn based meteorological records.<br> The prediction models used in this paper can be classified into<br> model-driven approaches and data-driven approaches, according to<br> the different modeling methodologies. The model-driven approaches are based on crop mechanistic<br> modeling. They describe crop growth in interaction with their<br> environment as dynamical systems. But the calibration process of<br> the dynamic system comes up with much difficulty, because it<br> turns out to be a multidimensional non-convex optimization problem.<br> An original contribution of this paper is to propose a statistical<br> methodology, Multi-Scenarios Parameters Estimation (MSPE), for the<br> parametrization of potentially complex mechanistic models from a<br> new type of datasets (climatic data, final yield in many situations).<br> It is tested with CORNFLO, a crop model for maize growth. On the other hand, the data-driven approach for yield prediction<br> is free of the complex biophysical process. But it has some strict<br> requirements about the dataset.<br> A second contribution of the paper is the comparison of these<br> model-driven methods with classical data-driven methods. For this<br> purpose, we consider two classes of regression methods, methods<br> derived from linear regression (Ridge and Lasso Regression, Principal<br> Components Regression or Partial Least Squares Regression) and<br> machine learning methods (Random Forest, k-Nearest Neighbor,<br> Artificial Neural Network and SVM regression).<br> The dataset consists of 720 records of corn yield at county scale<br> provided by the United States Department of Agriculture (USDA) and<br> the associated climatic data. A 5-folds cross-validation process and<br> two accuracy metrics: root mean square error of prediction(RMSEP),<br> mean absolute error of prediction(MAEP) were used to evaluate the<br> crop prediction capacity.<br> The results show that among the data-driven approaches, Random<br> Forest is the most robust and generally achieves the best prediction<br> error (MAEP 4.27%). It also outperforms our model-driven approach<br> (MAEP 6.11%). However, the method to calibrate the mechanistic<br> model from dataset easy to access offers several side-perspectives.<br> The mechanistic model can potentially help to underline the stresses<br> suffered by the crop or to identify the biological parameters of interest<br> for breeding purposes. For this reason, an interesting perspective is<br> to combine these two types of approaches.
BibTeX
@inproceedings{chen2018model,
author = {Xiangtuo Chen and Paul-Henry Cournède},
title = {{Model-Driven And Data-Driven Approaches For Crop Yield Prediction: Analysis And Comparison}},
booktitle = {Zenodo},
year = {2018},
doi = {10.5281/zenodo.1316506},
}Loading related content...