Machine Learning-based Pre-season Crop Type Mapping: A Comparative Study
Machine Learning-based Pre-season Crop Type Mapping: A Comparative Study
复制标题
基于机器学习的季前作物类型测绘：比较研究
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
L. Di
中科院分区：
文献类型：
--
作者：
Alexander Yao;L. Di
可靠的作物类型信息对农业决策至关重要。然而，后季节作物类型信息不能支持季节估计和监测。相反，季前作物测绘可以为农业产量和供应链提供早期预警，以减少贸易紧张局势和农业风险。本研究基于历史作物类型数据，使用多种机器学习算法对季前作物类型图进行分析和预测。我们评估了三种机器学习算法，包括随机森林算法、极端梯度提升算法和朴素贝叶斯算法作为实验开发的预测模型。结果表明，在大数据集的情况下，机器学习算法在复杂的种植模式下比简单的种植模式具有更高的准确率。多种机器学习算法的比较研究表明，更复杂的算法，如随机森林算法，可以高效、低成本地生成合理的季前作物地图。
Reliable crop type information is crucial for decision-making in agriculture. However, post-season crop type information cannot support in-season estimation and monitoring. Instead, pre-season crop mapping can provide early warnings for agricultural yield and supply chains to reduce trade tension and agriculture risk. This research analyzed and predicted pre-season crop type maps using multiple machine learning algorithms based on historical crop type data. We evaluated three machine learning algorithms including Random Forest, Extreme Gradient Boosting, and Naïve Bayes as prediction models for experimental development. The results show that the machine learning algorithm with a large dataset has higher accuracy in complex cropping patterns than simple cropping patterns. The comparative study of multiple machine learning algorithms shows that more complex algorithms like Random Forest could produce reasonable pre-season crop maps in an efficient and low-cost way.