Predicting brown planthopper populations in diverse agro-climatic zones of India: a comparative study of statistical and machine learning models

The brown planthopper (BPH) causes direct damage to rice plants by sucking sap and indirectly transmits viral diseases. This study aimed to forecast BPH populations using a statistical model (SARIMAX) and machine learning models (MLP and LSTM), incorporating biotic (BPH count) and abiotic (climatic parameters) factors as input predictors. Among the models, LSTM outperformed MLP and SARIMAX, achieving higher accuracy (85.4%–98.5%), but requiring greater computational resources. In machine learning models, the previous week’s BPH count and current week’s total rainfall were the most significant features for predicting BPH populations across agro-climatic zones of India.