Personalized Fashion Advice - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 213

successful provision of fashion recommendations. Section 8.4 covers a detailed
explanation of the mentioned online shopping service for men with an overview
of the starting point for our approach. This section also covers the description of the
data used in the following sections. In Sect. 8.5 , we present the theoretical foundation
and explanation of our proposed approach. The preliminary evaluation in Sect. 8.6
presents first results which are also discussed. Section 8.7 presents an outlook and
outlines directions for future research.
8.2 Recommender Technologies in e-Commerce
Recommendation methods used in e-commerce, from most-popular to hybrid rec-
ommendations, are a common tool to assist people in finding suitable items to buy.
A well-known and often-used approach is collaborative filtering, that uses groups of
similar users, based on the purchase history, to recommend items. In the following
section, we give a short overview of recommendation systems used in e-commerce
applications and take a closer look at the domain of clothing recommendations. We
will show that there is a difference in recommending brown goods or clothes.
Studies show that the process of actual buying clothes is heavily related to the
mood of the user [ 38 ]. This means that having recommendations, the decision if a
piece of cloth is bought or not is out of control for the algorithm.
Other studies showed that user preferences for clothes depend on their physical
features [ 40 , 45 ]. Raunio identified different physical features of clothes including
skin response, size and shape of the clothes, thermal comfort, and fit (looseness
and over-sized) revealing levels and visual features as important for the selection of
clothes.
Delong et al. [ 14 ] found that preferences are composed of two parts: cognitive
and affective. Affective preferences refer to emotions and mood of the user. The cog-
nitive preferences are again referring to physical features such as product attributes,
esthetic, and social attributes. The preference for product features are either extrinsic
(e.g., price or brand name) or intrinsic (e.g., style, color, fabric, care, fit and quality)
but they can differ based on the category of clothes, e.g., casual wear [ 11 , 12 , 17 , 32 ].
In the next section, we will present current approaches to recommendation in the
area of e-commerce, with a focus on clothes recommendations.
8.2.1 Current Approaches for RS in e-Commerce
Since the dawn of e-commerce applications in the WWW, recommendation sys-
tems (RS) have become an important tool to help users cope with the Information
Overload problem and to help shop owners sell more items [ 41 ]. One well-known
example is the amazon.com RS where users get “Other people who bought this also
bought…” recommendations [ 31 ]. Current RS frequently use memory-based and