Controlled Temporal Feature Ablation and Explainable Ensemble Learning for Next- Hour Residential Electricity Forecasting

Notice

This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.

Year : 2026 | Volume : 16 | 03 | Page :
By

Priyanka Mitesh Patel,

  1. PhD scholar, Department of Computer Science, Sabarmati University, Gujarat, India

Abstract

Short-term residential load forecasting is increasingly used in demand response, home energy management, and data-driven grid operation, yet reported gains are often difficult to attribute to either the forecasting algorithm or the temporal representation supplied to it. This study presents a controlled, leakage-aware evaluation of temporal feature engineering for one-hour-ahead household electricity forecasting. The public UCI Individual Household Electric Power Consumption dataset was aggregated to hourly resolution, causally imputed using past observations only, and transformed into five feature families: electrical measurements, calendar variables, autoregressive lags, rolling statistics, and cyclical encodings. Nine predefined feature configurations (E1-E9) were evaluated with Linear Regression, Random Forest, Extra Trees, and Gradient Boosting under a fixed chronological 80/20 holdout comprising 27,536 training and 6,884 test observations. The best single configuration, E9 with all 24 features and Extra Trees, achieved MAE=0.3203 kW, RMSE=0.4587 kW, sMAPE=33.80%, WAPE=32.37%, and R²=0.6047. Relative to the current-hour persistence baseline, RMSE decreased by 22.25% (moving-block-bootstrap 95% CI: 0.1193-0.1442 kW absolute improvement), and a HAC-adjusted forecast-loss comparison was significant at p<0.001. Calendar and cyclical representations produced the largest early gains, while Gradient Boosting delivered the best average performance across the nine feature settings. SHAP analysis identified global current intensity, sub-metering load, and hour-of-day cyclic terms as the dominant predictive drivers. The findings show that disciplined temporal representation can be at least as consequential as model choice and provide a reproducible benchmark for interpretable residential load forecasting.

Keywords: residential load forecasting; temporal feature engineering; Extra Trees; Gradient Boosting; explainable AI; SHAP; time-series evaluation; smart energy

How to cite this article: Priyanka Mitesh Patel. Controlled Temporal Feature Ablation and Explainable Ensemble Learning for Next- Hour Residential Electricity Forecasting. Trends in Electrical Engineering. 2026; 13(03):-.
How to cite this URL: Priyanka Mitesh Patel. Controlled Temporal Feature Ablation and Explainable Ensemble Learning for Next- Hour Residential Electricity Forecasting. Trends in Electrical Engineering. 2026; 13(03):-. Available from: https://journals.stmjournals.com/tee/article=2026/view=259666

References

  1. Biswal B, Deb S, Datta S, Ustun TS, Cali U. Review on smart grid load forecasting for smart energy management using machine learning and deep learning techniques. Energy Reports. 2024 Dec 1;12:3654-70.
  2. Baur L, Ditschuneit K, Schambach M, Kaymakci C, Wollmann T, Sauer A. Explainability and interpretability in electric load forecasting using machine learning techniques–a review. Energy and AI. 2024 May 1;16:100358.
  3. Kim TY, Cho SB. Predicting residential energy consumption using CNN-LSTM neural networks. Energy. 2019 Sep 1;182:72-81.
  4. J Mupenzi JN, Witarsyah D, Kusnadi A, Sunge AS. An explainable data-driven framework for short-term residential energy forecasting. Energy and Buildings. 2026 May 27:117713.
  5. Ermis S. Improving short-term HVAC load forecasting via temporal feature engineering: insights from the PLEIAData building dataset. Energy Informatics. 2026 Jun 8.
  6. Moon J, Maqsood M, So D, Baik SW, Rho S, Nam Y. Advancing ensemble learning techniques for residential building electricity consumption forecasting: Insight from explainable artificial intelligence. PloS one. 2024 Nov 14;19(11):e0307654.
  7. Berard GH. Individual household electric power consumption. UCI Machine Learning Repository. 2006.
  8. Hyndman RJ, Koehler AB. Another look at measures of forecast accuracy. International journal of forecasting. 2006 Oct 1;22(4):679-88.
  9. Bergmeir C, Benítez JM. On the use of cross-validation for time series predictor evaluation. Information Sciences. 2012 May 15;191:192-213. Page 16 of 16
  10. Diebold FX, Mariano RS. Comparing predictive accuracy. Journal of Business & economic statistics. 2002 Jan 1;20(1):134-44.
  11. Breiman L, Forests R. Machine Learning. Vol. 45. The Netherlands: Kluwer Academic Publishers. 2001:5-32.
  12. Geurts P, Ernst D, Wehenkel L. Extremely randomized trees [J]. Machine learning. 2006;63(1):3-42.
  13. Friedman JH. Greedy function approximation: a gradient boosting machine. Annals of statistics. 2001 Oct 1:1189- 232.
  14. Hao J, Ho TK. Machine learning made easy: a review of scikit-learn package in python programming language. Journal of educational and behavioral statistics. 2019 Jun;44(3):348-61.
  15. Abdel-Basset M, Hawash H, Sallam K, Askar SS, Abouhawwash M. STLF-Net: Two-stream deep network for short- term load forecasting in residential buildings. Journal of King Saud University Computer and Information Sciences. 2022 Jul;34(7):4296-311.
  16. Abdelfattah E, Bowlyn K. Application of Machine Learning Models on Individual Household Electric Power Consumption. In2023 IEEE World AI IoT Congress (AIIoT) 2023 Jun 7 (pp. 0143-0146). IEEE.
  17. Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Advances in neural information processing systems. 2017;30.
  18. Bhandary A, Dobariya V, Yenduri G, Jhaveri RH, Gochhait S, Benedetto F. Enhancing household energy consumption predictions through explainable AI frameworks. Ieee Access. 2024 Mar 4;12:36764-77.
  19. Eskandari H, Saadatmand H, Ramzan M, Mousapour M. Innovative framework for accurate and transparent forecasting of energy consumption: A fusion of feature selection and interpretable machine learning. Applied Energy. 2024 Jul 15;366:123314.
  20. Mubarak H, Stegen S, Bai F, Abdellatif A, Sanjari MJ. Enhancing interpretability in power management: A time- encoded household energy forecasting using hybrid deep learning model. Energy Conversion and Management. 2024 Sep 1;315:118795.
Ahead of Print Subscription Review Article
Volume 13
03
Received 11/09/2026
Accepted 13/09/2026
Published 09/10/2026
Publication Time 28 Days

Login

My IP

PlumX Metrics

Support