基于EEMD-LSTM模型的禽霍乱预测研究

Research on Prediction of Fowl Cholera Based on EEMD-LSTM Model

禽霍乱是一种由多杀性巴氏杆菌引起的接触性、败血性传染病,鸡、鸭和鹅等禽类均易感。由于目前我国广泛使用的禽霍乱弱毒疫苗和灭活疫苗副作用大、免疫期短且保护率低,免疫后的禽类仍有患病的风险。因此,对禽霍乱的监控尤为重要。利用MATLAB 2020b软件构建了基于集合经验模态分解(ensemble empirical mode decomposition,EEMD)和长短期记忆(long short-term memory,LSTM)模型的EEMD-LSTM组合模型的禽霍乱预测方法。利用2006年-2015年禽霍乱的发病数训练模型,预测2016年-2020年禽霍乱的发病数,并与实际发病数验证,然后通过计算实际发病数与预测发病数的线性回归系数R^(2)值和组内相关系数(intraclass correlation coefficient,ICC)值,分析实际发病数与预测发病数的一致性。结果显示,模型训练期的R^(2)值和ICC值分别为0.9935和0.997,其ICC值大于0.75并接近于1,表明该模型具有良好的预测能力,可用于预测禽霍乱的发病趋势;模型预测期的R^(2)值和ICC值分别为0.7507和0.825,其ICC值大于0.75,同时大于Landis和Koch的建议值0.80,表明该模型具有良好的可信度。该模型的建立可为禽霍乱的防控提供参考,同时也为该模型的其他应用研究提供理论依据。

Fowl cholera is a contagious and septic infectious disease caused by Pasteurella multocida,which is susceptible to poultry such as chickens, ducks and geese.Due to the large side effects, short immunization period and low protection rate of the attenuated fowl cholera vaccine and inactivated vaccine widely used in China, the immunized poultry are still at risk of disease.Therefore, the monitoring of fowl cholera is particularly important.In this study, MATLAB 2020 b software was used to construct an fowl cholera prediction method based on the EEMD-LSTM combined model of Ensemble Empirical Mode Decomposition(EEMD) and Long Short-Term Memory(LSTM) models.Using the training model for the number of fowl cholera from 2006 to 2015,the number of fowl cholera from 2016 to 2020 predicted, and verified with the actual number of cases, and then the linear regression coefficient R^(2) value of the actual number of cases and the predicted number of cases was calculated within the Intraclass Correlation Coefficient(ICC) value, the consistency between the actual number of cases and the number of predicted cases was analyzed.The results showed that the R^(2) value and ICC value of the model during training period were 0.9935 and 0.997,respectively, and the ICC value was greater than 0.75 and close to 1,which indicated that the model had good predictive ability and could be used to predict the incidence trend of fowl cholera.The R^(2) value and ICC value in the prediction period of the model are 0.7507 and 0.825,respectively, and the ICC value is greater than 0.75,which is also greater than the recommended value of Landis and Koch, 0.80,indicating that the model has good credibility.The model can provide a reference for the prevention and control of fowl cholera, and also provide a theoretical basis for other application research of the model.