Advances in Online Shopping Interfaces: Product Catalog Maps and Recommender Systems


Update content

Title Advances in Online Shopping Interfaces: Product Catalog Maps and Recommender Systems
Period 01 / 2008 - 05 / 2010
Status Completed
Dissertation Yes
Research number OND1328770
Data Supplier Website ERIM


In many product categories, for example real estate and electronics, a consumerhas to choose from a heterogenous range of products with a largeamount of product characteristics. When a customer wants to select sucha product online, he or she often has to make a selection based on a (limited)number of constraints on product characteristics. The products thatsatisfy these constraints are usually shown in a list. A disadvantage of thisapproach is that customers can find these constraints too strict and thatproduct characteristics can substitute each other.Another approach would be to let the customer describe an ideal productbased on her ideal values for the product characteristics. Then, products arechosen that are most similar to the product. The system that determineswhich products are suggested to the user is called a recommender system.The usual approach of presenting the products in an ordered list has the disadvantagethat no information is given on how similar the selected productsare to each other. For example, two products that have almost the samesimilarity to the ideal product can differ from the ideal product on a completelydifferent set of characteristics and thus differ a lot from each other.Therefore, an ideal system that shows products to customers should not onlybe based on the similarities of products to an ideal product, but also on themutual similarities of the selected products.In this research project, we will visualize products in a two dimensionalspace, such that the mutual similarities between all products will be preserved.We intend to develop two systems using these 2D spaces. The firstsystem is the graphical recommender system (GRS), that asks the user togive a description of her ideal product. Then, the GRS determines whichproducts should be suggested and presents them, together with the idealproduct specification, in a 2D space. The second system, the graphical shoppinginterface (GSI), can be used to navigate through the product space. Thesystem will suggest a set of products to the user in a two dimensional space,without asking the user for a product specification. When the user selectsa product, a new set of products will be suggested, based on the productthat was selected. In this way, the user can navigate towards the product heprefers.An important requirement for the graphical shopping interface will beits responsiveness and adaptiveness to the user: The representation of theproduct space must adapt to the user preferences. This adaption can either2be explicit, e.g. the user indicating characteristics are important to her, orimplicit, e.g. the system infers important characteristics from user behavior.The exact form of this adaptivity is part of this research project.The remainder of this proposal is structured as follows. In the nextsection, we describe a prototype of the graphical shopping interface. Then,in Section 3, we give a short description of recommender systems. Section 4presents the research questions. The methodology to be used to answer theresearch questions is described in Section 5. In Section 6, the relevance isgiven and Section 7 provides a tentative planning. Finally, in Section 8, wewill discuss the possibilities of cooperation.

Abstract (NL)

Gedurende de laatste twee decennia is het internet snel een belangrijk medium geworden om informatie te vinden, sociale contacten te onderhouden en om online te winkelen. Dit laatste heeft een aantal voordelen ten opzichte van traditioneel winkelen. Producten zijn vaak goedkoper via internet, internetwinkels bieden een breder assortiment en kan de consument op het internet winkelen op het tijdstip dat hij zelf wil en dat zonder het verlaten van zijn eigen stoel. Daarentegen hebben de huidige internetwinkels nog twee belangrijke nadelen ten opzichte van `echte' winkels: producten zijn vaak moeilijker te vinden en er is geen verkoper om de consument te adviseren. Kagie gaat in zijn proefschrift in op beide nadelen. Zo bedacht Kagie verschillende nieuwe gebruikersomgevingen voor online winkels die gebaseerd zijn op het representeren van producten op kaarten in plaats van in lijsten, zodat producten makkelijker te vinden zijn. Op deze kaarten staan producten die veel op elkaar lijken dicht bij elkaar. Om deze kaarten te maken, maakte Kagie gebruik van statistische methoden zoals meerdimensionale schaling. Om een oplossing te bieden voor het tweede nadeel combineerde hij deze kaarten met een aanbevelingssysteem, dat de gebruiker assisteert het beste product te vinden. Tevens introduceert hij een aanbevelingssysteem dat kan uitleggen waarom het een bepaald product aanbeveelt aan de gebruiker. De methodes die gebruikt worden in dit proefschrift kunnen volgens Kagie de basis vormen voor nieuwe veelbelovende online winkelomgeving.

Related organisations

Related people

Supervisor Prof.dr. P.J.F. Groenen
Co-supervisor Dr. M. van Wezel
Doctoral/PhD student Dr. M. Kagie


D16500 User interfaces, multimedia
D70000 Economics and Business Administration

Go to page top
Go back to contents
Go back to site navigation