Please use this identifier to cite or link to this item:
Title: Reducción de la dimensionalidad mediante métodos de selección de características en microarrays de ADN
Author: Maseda Tarin, Miguel
Director: Ventura Royo, Carles  
Tutor: Isern Alarcón, David
Keywords: dimensionality reduction
hybrid methods
feature selection
Issue Date: 2-Jan-2019
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: This document deals with the curse of dimensionality that we can find in datasets, to be specific, in those datasets where the number of features are counted as hundreds or thousands in each sample. We are looking for an improvement in the classification of our dataset through feature selection methods. The application of feature selection methods to datasets aims to reduce their dimensionality, with the intention of finding a feature subset that explains the problem in the appropriate way. With that in mind, we will use one benchmark dataset in the DNA microarrays studies, we will apply four feature selection methods so that we can verify the results with three different machine learning algorithms. We will use two filter methods (f-score and mRMR), one wrapper (SFS_forward) and, specially, a hybrid method, an adaptation of the work . We can observe the different advantages and issues offered by each of the feature selection methods and the different results that we obtain based on the machine learning method used for the classification task.
Language: Spanish
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
mmasedaTFG0119_Video_Presentación.mp4161.78 MBMP4View/Open
mmasedaTFG0119_Presentación.pptx2.57 MBMicrosoft Powerpoint XMLView/Open
TFG_basic_functions.ipynb31.72 kBUnknownView/Open
TFG_F_W_Methods.ipynb1.86 MBUnknownView/Open
TFG_Hybrid_Method.ipynb3.03 MBUnknownView/Open
mmasedaTFG0119memoria.pdfMemoria del TFG2.91 MBAdobe PDFView/Open
mmasedaTFG0119presentación.pdfPresentación en PDF del TFG2.5 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons