Disease Diagnosis of Dairy Cow by Deep Learning Based on Knowledge Graph and Transfer Learning
Meng Gao,Northeast Agricultural University College of Electrical Engineering and Information,Haodong Wang,Weizheng Shen,Zhongbin Su,Huihuan Liu,Yanling Yin,Yonggen Zhang,Yi Zhang,Harbin 150020 Dairy Association of Heilongjiang Province,Northeast Agricultural University College of Animal Science and Technology,East University of Heilongjiang College of Computer Science and Electrical Engineering,,,,,,,,
DOI: https://doi.org/10.7546/ijba.2021.25.1.000812
2021-03-01
International Journal Bioautomation
Abstract:In dairy herd management, it is significant and irreplaceable for veterinarians to make rapid and effective diagnosis of dairy cow diseases. Based on electronic medical records, deep learning (DL) has been widely used to support clinical decisions for humans. However, this method is rarely adopted in veterinary diagnosis. In addition, most DL models are driven by large datasets, failing to utilize the knowledge acquired by veterinarians in subjective experience, which is critical to disease diagnosis. To address these problems, this paper proposes a DL method for disease diagnosis of dairy cow: convolutional neural network (CNN) based on knowledge graph and transfer learning (KGTL_CNN). Firstly, the structural knowledge was extracted from a knowledge graph of dairy cow diseases, and treated as part of the inputs to the CNN based on knowledge graph (KG_CNN). Then, the model performance was enhanced through pre-training by transfer learning. To verify its performance, experiments were carried out on dairy cow clinical datasets. The results show that our model performed satisfactorily on disease diagnosis: the KG_CNN and KGTL_CNN achieved an F1-score of 85.87% and 86.77%, respectively, higher than that of typical CNN by 6.58% and 7.7%. The research results greatly promote the effective, fast, and automatic clinical diagnosis of dairy cow diseases.