This study examines the relationships among perceived safety, social influence, technology anxiety, user satisfaction, and behavioral intention in blended learning implementation, investigating whether technology anxiety can be interpreted as productive vigilance rather than merely a barrier to technology adoption. Based on the Transactional Theory of Stress and Coping, this research explores how institutional safeguards and social expectations influence technology anxiety among teachers in high-stakes educational settings. A quantitative cross-sectional approach was applied using purposive sampling involving 205 high school teachers in North Sumatra, Indonesia. Data were analyzed using PLS-SEM to assess the structural model and mediation effects. The results indicate that Perceived Safety (β=0.526, t=8.17, p<0.001) and Social Influence (β=0.232, t=3.60, p<0.001) positively influence Technology Anxiety, explaining 48.0% of its variance. Technology Anxiety positively affects User Satisfaction (β=0.774, t=17.40, p<0.001), which subsequently influences Behavioral Intention (β=0.503, t=4.92, p<0.001). The direct effect of Technology Anxiety on Behavioral Intention is insignificant (β=-0.167, t=1.63, p=0.104), while the indirect effect through User Satisfaction is significant (βindirect=0.389, p<0.001). These findings highlight that technology anxiety may represent heightened attentiveness associated with satisfaction rather than resistance, requiring further validation through longitudinal and experimental studies.
Rice prices as a primary food commodity in Indonesia often fluctuate, affecting food-price stability and public welfare. This study analyzes and forecasts Medium I and Medium II rice prices in Kudus City using ARIMA, Prophet, Hybrid ARIMA–Prophet, Random Forest, and benchmark models. Daily PIHPS price data from 2021–2025 were divided chronologically into 781 training observations from 2021–2023 and 523 testing observations from 2024–2025. Model performance was evaluated using MAE, RMSE, MAPE, R², and directional metrics. Hybrid ARIMA–Prophet achieved MAE of Rp369.68 and MAPE of 2.62% for Medium I, and MAE of Rp467.31 and MAPE of 3.38% for Medium II. However, simpler benchmark models achieved lower price-level errors than the hybrid model. Random Forest showed higher directional Accuracy than Hybrid ARIMA–Prophet, but its low Precision and F1-Score indicate that Accuracy should be interpreted cautiously. The directional analysis also showed a highly imbalanced distribution, with unchanged movements accounting for approximately 94% of test observations. The Diebold–Mariano test, applied only to the regression forecast errors of Hybrid ARIMA–Prophet and Random Forest, showed significant differences for both Medium I and Medium II (p < 0.001). Overall, the findings indicate that more complex models do not necessarily outperform simpler benchmarks for the studied rice-price series.
Oil palm block-level productivity (Yield Per Hectare/YPH) relies on complex agronomic, climatic, and operational interactions, yet existing predictive models often neglect internal estate conditions. We developed an XGBoost-SHAP framework using 1,351 block-year observations (2018–2024) from a single plantation in Central Kalimantan, Indonesia. Features include plant age, land area, rainfall, and a labor ratio functioning as a land-based workload proxy. The baseline random-split testing achieved R² = 0.6276, MAE = 2.2747 Tons/Ha, RMSE = 3.3303 Tons/Ha, and MAPE = 28.73%. However, evaluating the model under grouped block and time-based validation schemes revealed limited spatial and temporal transferability. SHAP analysis identified plant age (47.51%), land area (14.32%), annual rainfall (10.87%), and labor ratio (8.51%) as primary yield drivers, revealing non-linear interaction patterns. Rather than a fully autonomous tool, this framework provides a transparent interpretive heuristic and decision-support prototype to evaluate block-level operational strategies. By quantifying internal operational variables alongside agronomic factors, this study contributes valuable insights for precision agriculture, suggesting future work validate dashboard heuristics with estate managers and incorporate granular fertilization and harvesting logs to further enhance predictive robustness.
AI-based emotion recognition is gaining increasing attention in child-oriented digital environments, including digital play, AI toys, online learning, therapeutic platforms, and interactive child-computer systems. However, the translation of recognized emotions into responsible caregiver-facing early warnings remains insufficiently explored. This systematic literature review synthesizes technical and socio-technical perspectives on child emotion recognition, focusing on sensing modalities, AI approaches, datasets, alert interpretation, trust, privacy, ethics, governance, and caregiver acceptance. Following PRISMA 2020 guidelines, 400 records from Scopus and IEEE Xplore were screened, resulting in 32 studies included in the core synthesis. The findings reveal that facial-expression analysis and convolutional neural networks dominate current research, while child-specific datasets, multimodal learning, real-world validation, uncertainty communication, privacy-by-design, and caregiver-centered evaluation remain limited. This review proposes a socio-technical framework linking AI emotion inference with explainable alert translation, privacy-aware governance, and caregiver decision support. The proposed emotion-to-alert mechanism remains conceptual and requires empirical validation before practical deployment.
Accurate crop yield prediction is essential for supporting food-security assessment, agricultural planning, and policy-level decision-making. This study proposes a modular agroclimatic feature engineering framework for country-level crop yield prediction using XGBoost and LightGBM. The proposed framework integrates climate indicators, climate–pesticide interactions, temporal descriptors, composite agroclimatic indices, and nonlinear transformations to improve predictive representation while controlling data leakage. Experiments were conducted on 28,151 country–crop–year observations from 98 countries covering 1990–2013. To evaluate temporal and geographic generalization, models were assessed using time-based validation, Random KFold, GroupKFold by country, bootstrap confidence intervals, and held-out-country evaluation. Results show that LightGBM with the S3 feature configuration achieved the best temporal prediction performance, obtaining an R² of 0.9492 and RMSE of 21,257 hg/ha. However, held-out-country evaluation revealed lower transferability, with the best configuration achieving R² of 0.6694, highlighting the challenge of geographic generalization. The findings demonstrate that engineered agroclimatic features can significantly improve country-level crop yield prediction, but model performance depends strongly on the validation setting. This framework provides a reliable benchmark for leakage-controlled agricultural machine learning research rather than direct farm-level operational forecasting.
