Please use this identifier to cite or link to this item:

http://hdl.handle.net/10609/88365
Title: Anàlisi predictiu del volum d'aigua en embassaments catalans
Author: Font Marcé, Jordi
Director: Parada Medina, Raúl
Tutor: Casas Roma, Jordi
Keywords: water volume
artificial neural networks
time series analysis
Issue Date: 6-Jan-2019
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: This master's thesis is based upon the data science methodology in order to compare several predictive models of water level in Catalan reservoirs, with the purpose of being a decision-supporting tool for the efficient management of this natural resource. Previous studies argue that artificial neural networks are the most optimal predictive model to carry out analysis in this field. In addition, support vector machines (SVM) and the Random forest classifier are also included as such. The results obtained in the comparison of these three models have been analysed, based on the use of the same univariate time series as input and-output of the model or the incorporation of multiple features as input of the model. The analysis of the reservoirs in Sau and Baells show that SVM model, with multiple input features, is the most suitable model to carry out predictions of such data. The work developed has proven a challenge in the sense of going beyond the state-of-the-art prediction models analysed without diminishing the accuracy of dealing with daily data. It also opens up the multivariate analysis as a way forward, but with the need to address the question of which factors should be considered, a key issue of high complexity in the object of study in question.
Language: Catalan
URI: http://hdl.handle.net/10609/88365
Appears in Collections:Bachelor thesis, research projects, etc.

Share:
Export:
Files in This Item:
File Description SizeFormat 
jordifmTFM0119memòria.pdfPresentació del TFM8.81 MBAdobe PDFView/Open
jordifmTFM0119presentació.pdfMemòria del TFM1.09 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons