Vol. 12, Issue 4 (2023)

Author(s):
Baswaraj Biradar, Sunil A Kulkarni, Shobharani M, Sidramappa, Gnyanadev Bulla, Santosh Rathod, Naveena K, Gayathri Chitikela and Fakeerappa Arabhanvi

Abstract:
Legumes are a vital source of high-protein food and play a crucial role in nitrogen fixation in the atmosphere. However, their productivity is threatened by various lepidopteran pests. In this study, we aimed to develop a robust statistical model for predicting the pest population of soybean, pigeonpea, and chickpea, using climatological input parameters as influencing variables. To achieve this, we employed improved statistical and machine learning models, such as INGARCH, Random Forest, Support Vector Regression (SVR), and neural network (ANN), to predict pest populations in the North Eastern Transitional belts of Kalyan Karnataka. We conducted the study at ARS, examining soybean tobacco caterpillar incidence, pod borer incidence in pigeon pea and chickpea crops, using various crop varieties, including Soybean: JS 335, Pigeonpea: BSMR-736, and Chickpea-JG-11, over a 15-year period (2006 to 2020), incorporating historical weather data comprising rainfall, maximum and minimum temperature, and relative humidity (morning and evening). Our results demonstrate that the ANN model is a highly viable and effective alternative for modeling and predicting the incidence of lepidopteron pests based on time-series data. Moreover, the Diebold-Mariano test statistics confirm the superiority of the ANN models over INGARCH, SVM, and Random Forest models. It is expected that machine learning techniques will be extensively used in the future to model the count time series of various crop pests in other crops.