Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Subjective matching of products using Machine Learning

Åström, Joacim LU and Ekdahl, Alexander (2017) In LU-CS-EX 2017-29 EDA920 20171
Department of Computer Science
Abstract (Swedish)
When insurance companies calculate compensation for products no longer on the market, they need to identify an equal product for val- uation. Subjectively matching products poses a unique challenge with a many-to-many pairing, which is currently handled by an algorithm that requires day-to-day maintenance. This thesis investigates current methodologies and tools and presents unique machine learning meth- ods in an attempt to improve results and eliminate manual labor. By combining a multilayered neural network with cosine proximity com- parison, this thesis shows promising results in that a fully automated machine learning model can be implemented and replace the systems currently in use.
Popular Abstract
Insurance companies manually find replacements for thousands of products every day. Can a machine replace the manual labor while remaining faithful to the consumers’ desires?
Please use this url to cite or link to this publication:
author
Åström, Joacim LU and Ekdahl, Alexander
supervisor
organization
course
EDA920 20171
year
type
H3 - Professional qualifications (4 Years - )
subject
keywords
Machine learning, closest match, neural network, data processing
publication/series
LU-CS-EX 2017-29
report number
LU-CS-EX 2017-29
ISSN
1650-2884
language
English
id
8928874
date added to LUP
2017-12-06 12:40:35
date last changed
2017-12-06 12:40:35
@misc{8928874,
  abstract     = {{When insurance companies calculate compensation for products no longer on the market, they need to identify an equal product for val- uation. Subjectively matching products poses a unique challenge with a many-to-many pairing, which is currently handled by an algorithm that requires day-to-day maintenance. This thesis investigates current methodologies and tools and presents unique machine learning meth- ods in an attempt to improve results and eliminate manual labor. By combining a multilayered neural network with cosine proximity com- parison, this thesis shows promising results in that a fully automated machine learning model can be implemented and replace the systems currently in use.}},
  author       = {{Åström, Joacim and Ekdahl, Alexander}},
  issn         = {{1650-2884}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{LU-CS-EX 2017-29}},
  title        = {{Subjective matching of products using Machine Learning}},
  year         = {{2017}},
}