Garment recommendation in an e-shopping environment by using a Markov Chain and Complex Network integrated method

Abstract

In the e-commerce environment, website-based recommendation systems have been developed in order to help consumers without professional fashion-related knowledge to identify the most relevant garment products satisfying their specific personalized requirements in terms of fashion and functionalities. However, there are two main drawbacks in the existing recommendation systems: (1) the existing shopping data and/or human body data-based recommendations cannot effectively deal with fashion trends; (2) the shoppers’ experience and knowledge are rarely considered in these systems. To solve these two drawbacks, we propose in this paper a new garment recommendation mechanism using a Markov Chain and Complex Network integrated method. The Markov Chain method is employed as a suitable tool permitting one to extract fashion trends from a lower quantity of shopping data. The Complex Network method is used to formalize uncertain relations between human body shapes, garment fitting effects and garment features, provided by shopping experts. Compared with other existing recommendation methods, the proposed method is validated to be more robust and more interpretable owing to its capacity of handling body types and fashion trends.

Construction of womenswear matching expert system based on domain knowledge

“…In determining the number of subjects, we referred to previous articles of the same type. [40][41][42][43] Furthermore, in conjunction with this article, although the steps of a systematic evaluation are simple, it requires a high degree of cooperation of the subjects and carefully judging the category of each set of collocation (satisfactory or relevant). In conclusion, this study faced 20 randomly selected testers from the major groups.…”

“…5 The early concept of clothing matching emphasizes its artistry and aesthetics, elaborating reasonable garment collocation methods from the perspectives of skin color, body shape, color, occasion, accessories, season, etc. 6 At this stage, the construction of massive clothing image datasets has driven research in outfits toward computer vision, such as clothing pattern recognition, 7 fashion style identification, 8,9 and clothing text description. 10 However, current studies ignore the rich and valuable knowledge (rules) accumulated in the fashion domain, especially about clothing matching.…”

“…Besides, the existing fashion recommendation system fails to consider these experiences and knowledge. 6 In addition, Perrett et al 17 established rules between clothing color and objectively measured skin tones from an aesthetic and stylist perspective. Gan and Li 18 noted that the more uniform the style of different outfits, the higher harmony of the collocation.…”

Development of novel fashion design knowledge base by integrating conflict rule processing mechanism and its application in personalized fashion recommendations

“…The proposed knowledge base can be implemented an interactive, personalized design recommendation system. Zhang et al 44 put forth a new garment recommendation mechanism by using a Markov Chain and Complex Network integrated method. The proposed method can help consumers without professional fashion-related knowledge.…”

“…Zhang et al. 44 put forth a new garment recommendation mechanism by using a Markov Chain and Complex Network integrated method. The proposed method can help consumers without professional fashion-related knowledge.…”