Meiram, Daniyar (A.A. 2024/2025) Comparative analysis of XGBoost, LSTM and TCN for energy demand forecasting: accuracy, uncertainty and interpretability. Tesi di Laurea in Data visualization, Luiss Guido Carli, relatore Blerina Sinaimeri, pp. 67. [Master's Degree Thesis]

Preview

PDF (Full text) Download (1MB) | Preview

Abstract/Index

Literature review and theoretical foundations. Traditional energy demand forecasting methods. The rise of neural networks in time series forecasting. Uncertainty quantification in machine learning. Interpretability and explainable artificial intelligence. Methodology and experimental design. Overall research framework. Data description and sources. Data preprocessing and feature engineering. Model selection. Accuracy metrics. Uncertainty quantification methods. Interpretability methods. Software and hardware environment. Empirical analysis I–forecasting accuracy. Benchmarking experiments. Architecture comparison. Impact of feature sets. Hyperparameter sensitivity. Empirical analysis II–uncertainty quantification. Implementing UQ methods. Evaluation of uncertainty estimates. Uncertainty sources. Accuracy-uncertainty trade-off. Empirical analysis III–model interpretability. Global interpretability. Local interpretability. Temporal pattern discovery. Linking interpretability to accuracy and uncertainty.

References

Bibliografia: pp. 61-63.

Thesis Type:	Master's Degree Thesis
Institution:	Luiss Guido Carli
Degree Program:	Master's Degree Programs > Master's Degree Program in Data Science e Management (LM-91)
Chair:	Data visualization
Thesis Supervisor:	Sinaimeri, Blerina
Thesis Co-Supervisor:	Laura, Luigi
Academic Year:	2024/2025
Session:	Extraordinary
Deposited by:	Alessandro Perfetti
Date Deposited:	14 Jul 2026 14:29
Last Modified:	14 Jul 2026 14:29
URI:	https://tesi.luiss.it/id/eprint/46397

Downloads

Repository Staff Only

View Item