Pesticide Repurposing with KGE
Introduction
We are the first to develop a technique for pesticide repurposing by constructing a pesticide-crop-disease knowledge graph and adopting KGE as well as link prediction. Firstly, we collect detailed data about pesticides, diseases, and crops from multiple information sources, including online resources, encyclopedias, and specialized literature. Secondly, we construct a detailed knowledge graph based on the dataset, reflecting direct links between pesticides, diseases, and crops. It also reveals intricate indirect relationships among these elements. Thirdly, we use this knowledge graph as the basis to train several KGE models and evaluate the model performance through link prediction experiments. The experimental results show that this method can effectively discover new applications of pesticides in managing various crop diseases. These results provide a solid scientific foundation for in-depth exploration of the multi-functionality of pesticides and optimization of their use strategies.
Dependencies
- CUDA 11.8
- Python 3.10 (or later)
- torch==2.4.0+cu118
- GNU Complier Collection >=8.1.0
- OpenKE-Pytorch
Data
we collect extensive information on pesticides, crops and their pest control from multiple channels. These sources include but are not limited to academic publications, professional websites, Wikipedia and other online resources.
The code can be downloaded from there .The test datasets of pesticide knowledge graph can be download from there. The training set will be made public when the paper is accepted.
Get started
build OpenKE-Pytorch first.
bash build.sh
Then, you can run the following command to train the model. Take TransE for example.
python main.py --model_name TransE --data_path data/ \
--opt_method Adam --alpha 0.001 --train_times 1000 \
--batch_size 1024 --use_gpu True --sampling_mode normal \
--save_path ckpt
Or you can run the following command to test the train ed model. Take TransE for example.
python main.py --model_name TransE --data_path data/ \
--use_gpu True -test True --load_path ckpt/TransE.ckpt
Results
The results of KGE models in pesticide knowledge graph are shown in the following table.
Pre-trained KGE models
The pre-trained KGE models are uploaded. You can download it.