멜론 수확과 크기 예측을 위한 최적의 생육모델 비교에 관한 연구

Abstract

Improving crop productivity and providing food security are currently emerging as important topics around the world. Among these, melon is a fruit loved by many consumers around the world, and predictions of its production and quality have great significance in the agricultural field. In particular, accurately predicting melon growth is an important factor in increasing agricultural economics and productivity. The purpose of this study is to utilize various machine learning techniques to accurately predict melon growth. The study was conducted based on melon growth and greenhouse environment data provided by the Jeollanam-do Agricultural Research and Extension Services. Initial data was collected in minutes, which were preprocessed and converted to daily averages. Additionally, because growth data were not measured at regular intervals, linear interpolation was used to fill in missing values. After completing data preprocessing and feature selection, learning was conducted using Linear Regression, Random Forest, XGBoost, LSTM, and GRU among several machine learning models. The performance of each model was evaluated by MSE and R² values. As a result, the random forest model showed the best performance. The brute-force method was used to select the optimal parameters for this model, and the MSE value of the random forest model using the optimal parameters derived through this was 2.33, showing high prediction accuracy. The results of this study are expected to significantly improve agricultural efficiency and productivity, and in the future, we plan to conduct research to further improve the accuracy of the prediction model by including more diverse variables and utilizing more data.