Diagnosis of dairy cow diseases by knowledge-driven deep learning based on the text reports of illness state

Abstract

Expert system is the most commonly used method for auxiliary diagnosis of dairy cow diseases, which is complex to build and usually difficult for non-professional farmers to operate. Moreover, it cannot discover the implicit knowledge hidden in the observed symptoms. To address these problems, we proposed a knowledge-driven deep learning model for efficient diagnosis of dairy cow diseases. The model first selected the explicit features from the text reports of illness state. Then, our model employed a professional knowledge graph of dairy cow diseases for extracting implicit features. Both the explicit and implicit features were furtherly fed into a BiLSTM-CNN hybrid network to make a diagnosis. The experimental results showed that the F1 value of our model reached 94.89%, which was 9.53% and 2.49% higher than that of the best machine learning model XGBoost and the neural network model DE-CNN, respectively. Our model can accurately diagnose dairy cow diseases, especially those with similar or common symptoms, and it will provide a new idea for the auxiliary disease diagnosis of other animals.