Unveiling the features of successful ebay sellers of smartphones: a data mining sales predictive model

• Autor(a) principal: Silva, Ana Teresa Nunes Biscaia Correia da
• Data de Publicação: 2016
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/13595
• topic: Online sales; EBay sellers; DATA MINING; Smartphones; Marketing; Data analysis; Gestão internacional; Venda; Comércio eletrónico; Telemóvel; Análise de dados

EBay is one of the largest online retailing corporations worldwide, providing numerous ways for customer feedback on registered sellers. In accordance, with the advent of Web 2.0 and online shopping, an immensity of data is collected from manifold devices. This data is often unstructured, which inevitably asks for some form of further treatment that allows classification, discovery of patterns and trends or prediction of outcomes. That treatment implies the usage of increasingly complex and combined statistical tools as the size of datasets builds up. Nowadays, datasets may extend to several exabytes, which can be transformed into knowledge using adequate methods. The aim of the present study is to evaluate and analyse which and in what way seller and product attributes such as feedback ratings and price influence sales of smartphones on eBay using data mining framework and techniques. The methods used include SVM algorithms for modelling the sales of smartphones by eBay sellers combined with 10-fold cross-validation scheme which ensured model robustness and employment of metrics MAE, RAE and NMAE for the sake of gauging prediction accuracy followed by sensitivity analysis in order to assess the influence of individual features on sales. The methods were considered effective for both modelling evaluation and knowledge extraction reaching positive results although with some discrepancies between different prediction accuracy metrics. Lastly, it was discovered that the number of items in auction, average price and the variety of products available from a given seller were the most significant attributes, i.e., the largest contributors for sales.