Final Oral Defense of Naidas, Jezra Irene
The Department of Information Systems and Computer Science
invites you to the Final Oral Defense
of
Jezra Irene L. Naidas
MS DS
Multivariate Time Series Forecasting of Locational Marginal Prices in the Philippine Wholesale Electricity Spot Market
Adviser
John Noel C. Victorino, Ph.D.
Panelists
- Raphael B. Alampay, Ph.D.
- Mr. John Carlo Daquis
- Mr. Lorenz Ray Payonga
November 14, 2026, 6:00 PM - 7:30 PM (Philippine Standard Time)
Via Google Meet, Meeting link: https://meet.google.com/dna-ceqq-mdf?hs=224
Abstract
Locational marginal price (LMP) forecasting in the Philippine Wholesale Electricity Spot Market (WESM) is challenging due to pronounced price volatility, recurring intraday patterns, and short-lived price spikes. This study evaluates statistical and machine learning approaches for forecasting LMP across generator and load nodes in the Metro Manila grid using five-minute market data and weather, economic, and outage variables as candidate predictors. Feature selection excluded coal futures prices, foreign exchange rates, outage indicators, and pricing flags from the final models due to limited predictive value. The Multiple Seasonal-Trend decomposition using LOESS and Error, Trend, and Seasonality (MSTL+ETS) hybrid model effectively captured recurring intraday patterns but was less responsive to abrupt price movements. Two eXtreme Gradient Boosting (XGBoost) models complemented each other in forecasting normal and spike-related price movements, with recent LMP observations serving as important predictors in both. A regime-switching ensemble was then developed in which the normal-regime XGBoost forecasts were adjusted by a weighted deviation of the MSTL+ETS forecast from its trailing 12-observation moving average, while forecasts during detected spike regimes were taken directly from the spike-regime XGBoost model. The ensemble achieved an MAE of 1.18 PHP/kWh, RMSE of 2.97 PHP/kWh, and WAPE of 34.13% across the full test set, with the w = 75% weight providing strong overall performance. The results demonstrate the potential of combining recurring temporal structure with recent price information and regime-specific
forecasting to capture short-term price movements in the Philippine electricity market.

Latest Events
Lecture / Talk / Discussion
Navigating AOM Submissions: The Annual Meeting and Journals
Mon, 23 Nov 2026
Thesis / Dissertation Defense
Proposal Defense of Emrick James A. Panganiban, Christian Rey S. Paraan, and Ada Jemina T. Plata
Wed, 25 Nov 2026
Lecture / Talk / Discussion
An Indictment of Spanish Colonialism: The French Translation of Jose Rizal’s First Novel (1899)
Wed, 25 Nov 2026
Thesis / Dissertation Defense
Proposal Defense of Deanne Gabrielle D. Algenio, Vince Dexter Y. Ang, and Josh Tiffton M. Bolipata
Thu, 26 Nov 2026