Deep Learning for Multi-Source Data-Driven Crop Yield Prediction in Northeast China

Deep Learning for Multi-Source Data-Driven Crop Yield Prediction in Northeast China

2024 | Jian Lu, Jian Li, Hongkun Fu, Xuhui Tang, Zhao Liu, Hui Chen, Yue Sun, and Xiangyu Ning
This study evaluates the performance of the CNN-LSTM-Attention model in predicting crop yields for maize, rice, and soybeans in Northeast China, using multi-source data from 2014 to 2020. The model integrates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and an attention mechanism to process complex datasets and capture non-linear relationships. The study also explores the effectiveness of the Kernel Normalized Difference Vegetation Index (kNDVI) for yield prediction, comparing it with traditional indices like NDVI, EVI, and NDWI. The results demonstrate that the CNN-LSTM-Attention model significantly enhances yield prediction accuracy compared to traditional methods such as Random Forest (RF), XGBoost, and CNN. The study provides valuable insights for precision agriculture, contributing to global food security and sustainable agricultural development. Key contributions include the evaluation of the CNN-LSTM-Attention model, the introduction of kNDVI, and a comprehensive analysis of environmental and photosynthetically active variables on crop yields. The findings highlight the potential of advanced deep-learning technologies in improving yield prediction accuracy and food production strategies.This study evaluates the performance of the CNN-LSTM-Attention model in predicting crop yields for maize, rice, and soybeans in Northeast China, using multi-source data from 2014 to 2020. The model integrates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and an attention mechanism to process complex datasets and capture non-linear relationships. The study also explores the effectiveness of the Kernel Normalized Difference Vegetation Index (kNDVI) for yield prediction, comparing it with traditional indices like NDVI, EVI, and NDWI. The results demonstrate that the CNN-LSTM-Attention model significantly enhances yield prediction accuracy compared to traditional methods such as Random Forest (RF), XGBoost, and CNN. The study provides valuable insights for precision agriculture, contributing to global food security and sustainable agricultural development. Key contributions include the evaluation of the CNN-LSTM-Attention model, the introduction of kNDVI, and a comprehensive analysis of environmental and photosynthetically active variables on crop yields. The findings highlight the potential of advanced deep-learning technologies in improving yield prediction accuracy and food production strategies.
Reach us at info@study.space
Understanding Deep Learning for Multi-Source Data-Driven Crop Yield Prediction in Northeast China