Título: Intelligent Electronic Catalogs Based On Self-Organizing Maps: From Conception To Empirical Validation
Autores: Martins, Weber; Moraes, Lena L.; Nalini, Lauro E. G.; Tsukahara, Fernando P.
Resumo: This paper introduces and validates empirically a novel pro- posal of automatic personalized product catalogs for E-commerce based on topological ordering. The main goal is the construction of interactive intelligent interfaces (intelligent catalogs) that minimizes user search time and effort on choosing products in electronic catalogs on the Inter- net. The topological ordering is realized by Self-Organizing Maps (SOM) neural networks. The conception of personalized catalogs considers the degrees of importance (preference) of each product attribute which are provided by the user and define a weighted Euclidean distance metric. The proposed system achieves product clustering by employing users personal criteria and, therefore, enables the discovery of interesting prod- uct regions from the user perspective. Comparisons are conducted by recording user interactions in two virtual stores: traditional shop with hierarchical catalogs (clustering by product brands and models) and intelligent shop with topological organized catalogs (clustering by user preferences of product attributes and SOM). Descriptive and inferential statistical analyses of the collected data indicate promising results in fa- vor of the proposed system when users are familiar with the interface, since differences are significant (at 5%) in terms of search time (session duration) and selected product quality.
Palavras-chave: Self-Organizing Maps (SOM); electronic catalogs
Artigo em pdf: 60100066.pdf
Arquivo BibTex: 60100066.bib