Bilinear LSTM with Bayesian Gaussian Optimization for Predicting Tomato Plant Disease Using Meteorological Parameters

Climate change threatens agriculture; as a result, adaptation measures are required to withstand agricultural produce, reduce susceptibility, and improve the farm system's flexibility to climate change. Meteorological parameters like temperature and relative humidity play an essential role in the condition of disease occurrence in plants. We studied ARIMA, Prophet, and Long Short-Term Memory (LSTM) with stochastic gradient descent with momentum, RMSprop, and Adam optimizers to forecast the temperature and relative humidity. The work proposes a hybrid regression prediction model of Bilinear LSTM with Gaussian Bayesian optimization (BLSTM_bayOpt) for predicting disease in tomato plants based on weather parameters. From the six prediction models in this study, the performance of BLSTM_bayOpt in prediction with RMSE of 1.1573 and 5.5509, MAPE is 0.0556 and 0.0927, R2 is 0.9324 and 0.9475 for temperature and relative humidity, respectively. The proposed hybrid BLSTM_bayOpt model improved by 40.67%, with an MSE score for relative humidity prediction.