Angelina Balqis Khansa, Vivi Nur Wijayaningrum, Moch Zawaruddin Abdullah
Accurate weather information is crucial for agricultural planning, yet farmers in developing regions often face limitations due to the lack of localized and reliable forecasting tools. Existing forecasts are typically broad in scale and short-term in nature, making them less effective for guiding crop management decisions at the farm level. In citrus farming, where productivity is highly dependent on microclimatic factors, the lack of precise predictive systems increases the risk of yield reduction and inefficient resource allocation. To overcome these challenges, this study developed a weather prediction and classification system that integrates machine learning techniques with historical meteorological data from the Indonesian Meteorological, Climatological, and Geophysical Agency (BMKG). Support Vector Regression (SVR) was applied to predict four essential weather parameters, including relative humidity, average temperature, rainfall, and sunshine duration, while Classification and Regression Tree (CART) was employed to classify conditions into 'favorable' and 'unfavorable.' The methodology involved systematic data preprocessing, including parameter selection, imputation of missing values, augmentation via smoothing, normalization, and hyperparameter optimization, followed by functional testing to ensure usability. The experimental evaluation showed that SVR achieved strong predictive performance with RMSE values of 2.06 (RH_avg), 0.27 (Tavg), 4.53 (RR), and 0.68 (ss). CART provided excellent classification results with an accuracy of 0.9854, precision of 0.9855, recall of 0.9911, and an F1-score of 0.9882. These results demonstrate that the proposed system is both accurate and reliable, offering a practical decision-support tool to help farmers optimize citrus crop management under varying weather conditions. © 2025 IEEE.
Politeknik Negeri Malang, Department of Information Technology, Malang, Indonesia