RICE CROP YIELD PREDICTION USING HIERARCHICAL TIME SERIES FORECASTING WITH ENSEMBLE TECHNIQUES IN TELANGANA REGION

RICE CROP YIELD PREDICTION

Abstract

Agriculture is the main source of income especially rice production for the great majority of people. In agriculture, accurate rice crop yield forecasting is crucial for making various crop production-related policy decisions that will assure the supply of food. Taking into account, many works concentrating on machine learning and ensemble based techniques; however, it has several limitations such as increased complexity, overfitting and underfitting, struggle to generalize effectively. Therefore, to overcome the limitations this research proposed a novel rice crop yield prediction using hierarchical time series forecasting with ensemble techniques in Telangana region. In the initial stage, the collected dataset is pre-processed to improve the accuracy of the prediction in which the missing values in imputed by k-NN imputation method and normalization used to normalize the input data. Then, this pre-processed input data is start by organizing the crop yield data into a hierarchical structure by using grid based ARIMA model. Moreover, this research utilizing XGBoosting as a meta-model and train the meta-model using the new dataset. As a result, by combining the ARIMA-based Hierarchical time series forecasting model with a stacking ensemble provides a higher accuracy.