Utilizing Public Data for Annual Crop Yield Prediction: A Possibility Study

초록

In this study, the development of the regression model exclusively utilized open-access data. The dataset comprised environmental data, cultivation area data, and yield data. The environmental data was sourced from the Korea Meteorological Administration, while the cultivation area and yield data were obtained from the statistical yearbooks of each major production region. Fifteen years of monthly environmental datasets (from 2008 to 2022) were compiled, comprising variables such as temperature, humidity, precipitation, sunshine duration, and dew point, among others (totaling 30 columns). Given the structure of the monthly datasets, the yield data was aggregated on an annual basis. As a pre-treatment step, the twelve months of environmental data were grouped to correspond with the annual yield data. For the development of the regression model, CNN and stacking method were employed. Their performance was evaluated with different metrics and compared for determine the optimal regression model for predicting the annual yield of Korean wheat. © 2025 ASABE. All rights reserved.

키워드