Maskininlärning för klassificeringsalgoritmer

Examensarbete för kandidatexamen

Please use this identifier to cite or link to this item:
Download file(s):
File Description SizeFormat 
257375.pdfFulltext1.44 MBAdobe PDFView/Open
Type: Examensarbete för kandidatexamen
Bachelor Thesis
Title: Maskininlärning för klassificeringsalgoritmer
Authors: Gedda, Rebecca
Gullbrandson, Martin
Ivarsson, Aron
Yuan, Wenjin
Abstract: This report examines four basic machine learning algorithms for classification problems: linear regression, the Precptron-algorithm, the Winnow-algorithm and a type of artificial neural network called multi-layer Perceptron. The mathematical foundation of these algorithms are developed throughout the report and provides the reader with a better understanding of the algorithms. Furthermore the algorithms are implemented in MATLAB and the code is availible in appendix for reference. The proportion of missclassified data is used as the main indicator of an algorithms performance, however the precison and recall of the algorithms is also taken into account. Three different data sets are tested: two data sets are simulated in MATLAB and a third dermatologic data set is gathered from a public database. These data sets are used to evaluate the performance of the algorithms. The simulated sets are chosen such that they introduce certain difficulties for the classification of the data. Finally, two performance boosting algorithms are implemented and tested for each classification algorithm. The result of using the boosting algorithms are compared to the results of not using them. The median of the proportion of missclassified examples is for both simulated data sets 0 % for all algorithms. Moreover the four algorithms achieves a median error of 2:78 % for the dermatologic data set even if some spread and differences between the algorithms are noted.
Keywords: Grundläggande vetenskaper;Matematik;Basic Sciences;Mathematics
Issue Date: 2019
Publisher: Chalmers tekniska högskola / Institutionen för matematiska vetenskaper
Chalmers University of Technology / Department of Mathematical Sciences
Collection:Examensarbeten för kandidatexamen // Bachelor Theses

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.