Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Optimizing Product Assortments with Unknown Historical Transaction Data Using Nonparametric Choice Modeling and Random Forest Classification
KTH, School of Electrical Engineering and Computer Science (EECS).
KTH, School of Electrical Engineering and Computer Science (EECS).
2019 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Abstract [en]

Assortment optimization is a crucial problem for many firms who need to make decisions on which products to stock in their stores in order to maximize revenues. Optimizing assortments usually entails fitting choice models to historical data. To a large extent, this becomes a problem of understanding consumer behavior. In this paper, a two-step method is proposed for optimizing assortments for stores where there was no known sales data. First, training data was generated by optimizing assortments, using a non parametric choice model, on similar stores for which data is available. Then this training data isused to develop a series of random forest models which given parameters for a store can generate an optimal assortment. The features used in the random forest models were chosen based on consumer behavior theory and consisted of geographical and financial features as well as features regarding the composition of the stores. The data used in this report was provided by a major Swedish print distributor with over 1000 stores and 2500 products. The results presented in this paper show that this method outperforms the baseline in all cases studied. Furthermore, it was determined that geographic features are the essential type of features for the models to determine the optimal assortments for stores.

Abstract [sv]

Produktsortimentsoptimering är ett centralt problem för många företag som måste ta beslut om vilka produkter de ska lagerhålla för att maximera sin vinst. Att optimera produktsortiment brukar ofta innebära att träna valmodeller på historisk data. Detta blir ofta en fråga om att förstå konsumenters beteende. I denna uppsats presenteras en tvåstegs metod för att optimiera produktsortiment utan historisk data. I det första steget optimeras sortimentet med hjälp av en icke-parametrisk valmodell på liknande butiker där data finns tillgängligt. Sedan utvecklas Random Forest modellermed de optimerade sortimenten som träningsdata. Givna en rad parameterar kan dessa modeller generera optimala sortiment. Parametrarna som användes i Random Forest modellerna valdes baserat på konsumentteori and bestod av geografiska och finansiella parametrar så väl som parameterar som beskrev butikernas sammansättning. Datan som användes tillhandahölls av ett svenskt företag inom tryckbranschen som har över 1000 butiker och 2500 produkter i sitt sortiment. Resultaten som presenterades i denna uppsats visar att metoden presterar bättre än baslinjen i alla fall som studerades. Utöver detta, så beslutas det att geografiska parametrar är de viktigaste parametrarna för modelerna att ta beslut angående de optimala sortimenten.

Place, publisher, year, edition, pages
2019. , p. 12
Series
TRITA-EECS-EX ; 2019:282
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-261636OAI: oai:DiVA.org:kth-261636DiVA, id: diva2:1359365
Examiners
Available from: 2019-11-07 Created: 2019-10-09 Last updated: 2019-11-07Bibliographically approved

Open Access in DiVA

fulltext(1087 kB)16 downloads
File information
File name FULLTEXT01.pdfFile size 1087 kBChecksum SHA-512
05fc99c01d0f7309bb32adc9a96ecfb0b92be27d0be6e611245bbc5b1e93ecf1263aaf1f19eac09359bf0baa19e2beb20724c73a92c5f9ae9bea67a9886e3fd8
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 16 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 31 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf