Uma nova abordagem para a criação de perfis de clientes rentáveis utilizando machine learning em ambiente de cloud computing

• Ano de defesa: 2018
• Autor(a) principal: Carvalho, Leandro da Silva
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Civil; UFRJ
• Palavras-chave em Português: Modelos RFMP; Aprendizado computacional; Computação em nuvem; Modelos RFM; Comportamento do consumidor; Classificação não supervisionada; CNPQ::ENGENHARIAS::ENGENHARIA CIVIL
• Link de acesso: http://hdl.handle.net/11422/13422

This work aims to present the creation of a new model, called RFMP (Recency – Frequency – Monetary – Profitability), as an alternative to the traditional RFM model to evaluate whether the inclusion of a new parameter P – associated with customer profitability – will have any impact on targeting a client database from an e-commerce site. Additionally, a new classification methodology was proposed to increase the confidence level of the generated models and improve the assertiveness between the clusters and their respective contents, instead of the traditional form used. Besides, it was also recommended to create three new measurement indices in order to overcome the lack of existing indicators capable of determining the consistency and quality of the clusters and models produced. All of this was accomplished through an empirical study, done using a machine learning platform from a cloud computing environment. Finally, with the results obtained, it was possible to demonstrate that there was a direct impact on the formation of the clusters produced, since the customer groups were differentiated not only by the monetary value, but also from their respective profitability, which allowed to determine that customers with the highest monetary values were not always the most profitable.