Predicting Structured and Volatile Time Series: A Machine Learning Approach to Electrical Power Consumption and Bitcoin Prices

George, Jisno (2025) Predicting Structured and Volatile Time Series: A Machine Learning Approach to Electrical Power Consumption and Bitcoin Prices. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (516kB) | Preview

Abstract

The paper examines how machine learning (ML) regression can be utilized, namely XGBoost, to predict the short-term residential electricity use based on a daily aggregated electricity consumption data with respect to 2017. The temporal patterns were reflected in covariates engineered in the dataset, including lagged consumption values, rolling averages, and day-of-week (e.g., variables). XGBoost had an MAPE rate of around 6.89 percent, and RMSE of 0.415 kWh revealing top-notch predictive accuracy and approximation with real-world values of consumption. The results of this performance highlight the suitability of tree-based ensemble to structured, seasonal data. To set a comparative background, the study replicated the Bitcoin price data, which is a non-stationary and volatile time series (20112021) using the approach of modeling. In this case, the predictive accuracy was lower (RMSE in the hundreds of USD), as is the case when market volatility, poor seasonality, and reliance on exogenous determinants such as regulatory change or market mood make it difficult to model. This comparison shows the significance of dataset properties in the establishment of performance of the model. Tree-based models perform well in stable and seasonal data such as power consumption but not in volatile data, which need an adapting or the hybrid model that includes external interventions.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Byrne, Brian UNSPECIFIED
Subjects:	H Social Sciences > HG Finance > Money > Digital currency > Cryptocurrencies T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Electricity Supply H Social Sciences > HG Finance > Fintech T Technology > T Technology (General) > Information Technology > Fintech Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in FinTech
Depositing User:	Ciara O'Brien
Date Deposited:	20 Aug 2026 10:49
Last Modified:	20 Aug 2026 11:37
URI:	https://norma.ncirl.ie/id/eprint/9570

Actions (login required)

View Item