ARIMAX—Artificial neural network hybrid model for predicting semilooper ( Chrysodeixis acuta ) incidence on soybean

Insect pest and weather relations analysed using statistical models empower crop pest management through their capacity to forewarn abundance or their damage during season of crop cultivation. Field datasets of eight seasons (2010–2017) of Maharashtra (India) used to study the influence of weather factors lagged by one week on soybean semilooper ( Chrysodeixis acuta ) pest status based on Kendall’s correlations revealed significant and positive influence of maximum temperature (MaxT), MaxT deviation, minimum temperature (MinT), MinT deviation and rainfall (RF) with relative humidity (RH) influence significantly negative. The multiple linear regression (MLR) indicated the significance of MaxT, MaxT deviation, MinT deviation and RF, however with the totality of factors (R 2 = 0.04) accounting only four per cent of variations in semilooper. Nonlinear models individually and in hybrid mode fitted and tested using datasets indicated the best performance of hybrid of autoregressive integrated moving average with exogenous variable (ARIMAX) with artificial neural network (ANN) i.e. ARIMAX-ANN over ANN or ARIMAX models. Performance of the non-linear models were evaluated based on statistical measures of mean square error (MSE) and root mean square error (RMSE). The proposed artificial intelligence based non-linear ARIMAX-ANN model has potential for forecasting current week’s semilooper incidence on soybean one week in advance based on weather factors of previous week across Maharashtra. Paper also covers in detail the framework and methodological approaches of models in addition to their testing using appropriate statistical measures.