Comparative Study of ARIMAX and BP Neural Network Models in the Prediction of Spodoptera Frugiperda Occurrence
Jumei Chang, Ying Lu, Quanyuan Xu, Hanrui Zhang · 2023
This study compares the performance of the ARIMAX (Autoregressive Integrated Moving Average model-X) and BP (Backpropagation) neural network models in predicting the quantity of Spodoptera Frugiperda in grassland. The Spodoptera Frugiperda is a crop pest that causes significant losses in agricultural production. Accurately predicting its quantity is crucial for implementing appropriate prevention and control measures. We collected quantity data of Spodoptera frugiperda from 2021 to 2022 in grassland and extracted several exogenous variables that might influence its quantity, such as temperature, humidity, and precipitation. Firstly, we employed the ARIMAX model for prediction, which combines autoregressive, moving average, and integrated methods, while also incorporating the adjustment of the quantity using exogenous variables. Next, we utilized the BP neural network model, which employs the backpropagation algorithm to learn and train the network, in order to capture the nonlinear relationships within the data. By comparing the performance of the two models in quantity prediction, we found that the ARIMAX model demonstrates outstanding performance in short-term forecasting. It accurately captures the trends and periodic variations in quantity. The research findings can provide a scientific basis for pest control in agricultural production and serve as a reference for further improvement of prediction models.