This paper proposes a hybrid, interpretable machine learning framework for the preliminary screening of soil macronutrients using Sentinel-2 and AgroLens data. This study aims not to replace laboratory analysis, but to test the feasibility of obtaining a useful proxy signal for estimating nitrogen (N), phosphorus (P), and potassium (K) content using a limited set of remote sensing and agricultural features. The developed pipeline includes data auditing, leakage control, feature engineering, train-only normalization, group-aware partitioning, baseline/SOTA model comparison, hybrid regression modeling, SHAP interpretation, and uncertainty assessment. The experiment used 4471 AgroLens observations and 126 features derived from Sentinel-2 spectral aggregates, vegetation indices, temporal characteristics, and crop-related parameters. The evaluation indicated that the proposed approach consistently improves forecasting quality relative to baseline models under reduced-input conditions. Linear relationships between target variables ranged from 0.14 to 0.17, while nonlinear relationships reached 0.23. SHAP analysis revealed significant contributions from vegetation indices, crop-specific interactions, and Sentinel-2 spectral channels. The findings support the applicability of the proposed framework for preliminary monitoring, prioritizing field surveys, and decision support in digital agriculture. Although an additional AgroLens control segment was used to assess the robustness of the study, the study did not include independent external validation of the data collected across different geographic or agro-climatic conditions.

This study proposes a method for probabilistic time-series forecasting on graphs incorporating interpolation (IA-GTPF)—a decision-tree-based ensemble approach that combines observation quality metrics, historical load dynamics, information on neighboring nodes, residual correction, quantile regression, and conformal calibration. The method’s performance was evaluated using the Open Power System Data Household dataset. Fixed time boundaries defined distinct periods for training, validation, calibration, and testing. Only information available at the time of forecasting was used as input, while graph weights and preprocessing parameters were determined solely based on the training set. A separate calibration period was used to adjust confidence interval boundaries without altering point forecasts. On the main test set, the IA-GTPF method demonstrated the following metrics for a 15-min forecast horizon: a Mean Absolute Error (MAE) of 0.9824 kW, a Root Mean Square Error (RMSE) of 2.5408 kW, and a Coefficient of Determination (R²) of 0.98947. For the 60-min horizon, the corresponding values were 1.3940 kW, 4.2017 kW, and 0.97122. The QuantileForest model showed a lower MAE for the 15-min horizon (0.9695 kW); however, IA-GTPF achieved the lowest RMSE and highest R² values. For the 60-min horizon, the IA-GTPF method demonstrated the lowest MAE and RMSE values among all evaluated models.