| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 919.53 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Todos os dispositivos mantêm registos das nossas decisões, escolhas e hábitos, aumentando assim a quantidade de dados associados ao indivíduo. Estes dados são um mundo de informação para o comércio e indústria, atribuindo uma lógica a dados que antes não tinham significado. De acordo com este problema surgiu a temática para este projecto, uma Solução de Recomendação de Conteúdos Personalizados que permitirá, de acordo com os comportamentos do cliente registados, direccionar não só conteúdos disponíveis no serviço em questão, como também campanhas mais adequadas ao perfil comportamental do cliente. Esta Solução terá como base dados reais provenientes de um negócio de telecomunicações e a sua correspondência com os dados do IMDb. Na solução apresentada, utilizámos dois algoritmos de Aprendizagem Automática com o intuito de identificar grupos de clientes com os mesmos comportamentos. Estes grupos permitiram estabelecer três cenários distintos de recomendação com base nas features escolhidas. Para cada um dos cenários, foram recomendados conteúdos utilizando uma recomendação híbrida, neste caso um conjunto de três tipos diferentes de recomendação. Com a Solução desenvolvida, conseguimos obter recomendações apropriadas para 90% dos clientes da amostra utilizada. Também foi possível identificar através dos comportamentos destes mesmos clientes, que tipo de campanhas seriam as mais adequadas. Tudo isto influenciará a experiência do cliente para como serviço podendo motivar à sua fidelização para com o negócio.
All devices keep a record of our decisions, choices and habits, raising data volume associated with each person. Such data is extremely valuable to business and industries, providing meaning to information that before didn’t have any meaning. On a Solution for Personalized Content Recommendation, that will, from registered customer behavior, not only recommend pertinent visualization options but also provide valuable input for marketing campaigns based on the customer profile. The data used in this solution is originated from the real world deployment of a telecommunications company, and was matched with IMDb data entries. For such achievement, in the presented solution, we used two Machine Learning algorithms in order to identify customers groups with the same behavior. This groups aim to establish three different recommendation scenarios based on the choosen features. For each one of the scenarios, the content was also recommended using an hybrid recommendation, in this case a set of three different recommendation types. With the developed solution, we were able to obtain appropriate recommendations for 90% of the clients in the used sample. It was also possible to identify through these customer behaviors, which type of campaigns would be the most appropriate. All of this will influence the customer experience in the service being able to motivate its loyalty to the business.
All devices keep a record of our decisions, choices and habits, raising data volume associated with each person. Such data is extremely valuable to business and industries, providing meaning to information that before didn’t have any meaning. On a Solution for Personalized Content Recommendation, that will, from registered customer behavior, not only recommend pertinent visualization options but also provide valuable input for marketing campaigns based on the customer profile. The data used in this solution is originated from the real world deployment of a telecommunications company, and was matched with IMDb data entries. For such achievement, in the presented solution, we used two Machine Learning algorithms in order to identify customers groups with the same behavior. This groups aim to establish three different recommendation scenarios based on the choosen features. For each one of the scenarios, the content was also recommended using an hybrid recommendation, in this case a set of three different recommendation types. With the developed solution, we were able to obtain appropriate recommendations for 90% of the clients in the used sample. It was also possible to identify through these customer behaviors, which type of campaigns would be the most appropriate. All of this will influence the customer experience in the service being able to motivate its loyalty to the business.
Descrição
Tese de mestrado, Informática, Universidade de Lisboa, Faculdade de Ciências, 2019
Palavras-chave
Aprendizagem Automática Prospeccão de Dados Segmentação de Clientes Sistemas de Recomendação Teses de mestrado - 2019
