Incidence of Yellow stem borer (Scirpophaga incertulas)(YSB) on Rice (Oryza sativa L.)atChinsurah, West Bengal, India is modelled based on field data sets generated during six kharif seasons [2011-20]. The weather variables considered are maximum & minimum temperature (MaxT & MinT) (0C), morning and evening humidity (RHM & RHE) (%), sunshine hours (SS) (hr/d), wind velocity (Wind) (km/hr), total rainfall (RF) (mm) and rainy days (RD).Long Short-Term Memory (LSTM)networks, which arecapable of learning long-term temporal dependencies, are used to overcome the limitations oftraditional machine learning techniques. The results indicate that LSTMand Gated Recurrent Unit (GRU) models, although morecomputationally expensive, provide a more accurate solution for pest prediction compared withother methods. Correlation analyses indicate significant positive influence of maximum and minimum temperature on YSB. An empirical comparison of the above models is carried out based on root mean square error (RMSE) and mean square error (MSE). It is observed that, for YSB, the MSE and RMSE values of LSTM and GRU are less as compared to other competing models. Diebold-Mariano (D-M) test was applied for comparison of forecasting performance among the applied models. It is observed that, in the studied pest, predictive accuracy ofLSTM is higher than that of other models. The analysis is carried out using R package.