| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.23 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Atualmente, devido ao desenvolvimento das tecnologias de informação e computação e consequente melhoria do acesso à internet, verifica-se uma maior utilização das plataformas de comércio eletrónico para a compra de produtos. Contudo, a forte adesão das pessoas a este tipo de plataformas originou dificuldades para as empresas que criam estas plataformas. Um dos principais problemas é a grande quantidade de produtos disponíveis o que dificulta os utilizadores encontrarem os produtos que pretendem. No caso da Netflix, por exemplo, “caso um utilizador, não encontre algo que lhe interesse após um período de 60 a 90 segundos de navegação, a probabilidade de o mesmo não aderir ao serviço aumenta consideravelmente”. De forma a colmatar esse problema, foram criados os sistemas de recomendação, que visam ajudar os utilizadores no processo de compra, através da sugestão de produtos com base em diversos fatores. Este documento analisa algumas das arquiteturas utilizadas pelas plataformas de comércio eletrónico, diferentes técnicas de recomendação e os problemas associados a cada uma delas, diferentes métodos e métricas de avaliação de sistemas de recomendação. Além disso, apresenta alguns exemplos de sistemas de recomendação. Esta dissertação culminou na proposta de um sistema de recomendação para uma plataforma de comércio eletrónico. Os resultados obtidos foram positivos uma vez que o sistema criado passou nos testes de integração, os três algoritmos implementados obtiveram uma taxa de sucesso superior à da baseline1 , tendo os testes de carga realizados à API sido melhores do que o esperado.
(1 Baseline: valor mínimo utilizado para avaliar a qualidade do sistema de recomendação.)
Nowadays, due to the development of technology and internet access there is a greater use of e-commerce platforms for the purchase of products. The increased use of this platforms created several management challenges for these companies. One of the main problems is the large number of products available which makes it difficult for users to find the products they want. In the case of Netflix, for example, "if a user does not find something that appeals to him after a period of 60 to 90 seconds of browsing, the likelihood that he will not join the service increases considerably." In order to overcome this problem, recommendation systems have been created, to help users in the purchase process, by suggesting products based on several factors. This document analyses some of the architectures used by e-commerce platforms, the different recommendation techniques, problems associated with each of them, different methodologies and metrics for evaluating recommendation systems. In addition, it presents some examples of recommendation systems. This dissertation lead to the proposal of a recommendation system for an electronic commerce platform. The obtained results were positive given the success of the integration tests; the three implemented algorithms had a higher success rate than the baseline2 ; and the performed API load tests were better than expected. (2 Baseline: a minimum value used for evaluating the quality of the recommendation system.)
Nowadays, due to the development of technology and internet access there is a greater use of e-commerce platforms for the purchase of products. The increased use of this platforms created several management challenges for these companies. One of the main problems is the large number of products available which makes it difficult for users to find the products they want. In the case of Netflix, for example, "if a user does not find something that appeals to him after a period of 60 to 90 seconds of browsing, the likelihood that he will not join the service increases considerably." In order to overcome this problem, recommendation systems have been created, to help users in the purchase process, by suggesting products based on several factors. This document analyses some of the architectures used by e-commerce platforms, the different recommendation techniques, problems associated with each of them, different methodologies and metrics for evaluating recommendation systems. In addition, it presents some examples of recommendation systems. This dissertation lead to the proposal of a recommendation system for an electronic commerce platform. The obtained results were positive given the success of the integration tests; the three implemented algorithms had a higher success rate than the baseline2 ; and the performed API load tests were better than expected. (2 Baseline: a minimum value used for evaluating the quality of the recommendation system.)
Description
Keywords
Sistema de recomendação Comércio eletrónico Regras associativas Filtragem colaborativa Recommendation system e-commerce Association rules Collaborative filtering