| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.64 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Os sistemas de recomendação (RS) são utilizados em vários setores para prever, filtrar ou dar
recomendações aos utilizadores com base em comportamentos, padrões ou preferências anteriores.
Porém, quando é necessário fazer essas previsões para grupos, em que vários utilizadores com
preferências diferentes estão envolvidos, os RS tradicionais têm mais dificuldade em fornecer
recomendações que atendam às suas necessidades. Os sistemas de recomendação para grupos
(GRS) procuram resolver este problema ao fornecer recomendações personalizadas, geralmente
com base nas preferências agregadas de cada membro do grupo. A área do turismo é um exemplo
de uma área em que este tipo de recomendações é complexa. Com o objetivo de melhorar as
sugestões dadas aos grupos turísticos, este projeto sugeriu a integração de regras de associação,
através do algoritmo Apriori, no atual Microserviço Multiagente (MAMS) de um protótipo de GRS
de turismo, Grouplanner.
O algoritmo Apriori é uma importante ferramenta de mineração de regras de associação para
identificar padrões e associações entre objetos num conjunto de dados e pode ser usado para
descobrir associações entre locais de interesse (POI) frequentemente visitados, examinando as
preferências e experiências de viagem dos membros do grupo.
Os principais objetivos do projeto foram melhorar a seleção de POI, incentivar uma
abordagem mais personalizada às recomendações e modificar dinamicamente as recomendações
de acordo com as preferências individuais dos membros dos grupos, através de previsões
resultantes do algoritmo Apriori, integrando-o num dos microserviços desenvolvidos em .NET e
alojados no Azure, o MAMS, existente no protótipo de GRS.
Em particular, foi desenvolvido um algoritmo Apriori personalizado, em C#, assim como a
geração das regras de associação em conformidade com o sistema multiagente. Seguidamente,
foram conduzidos testes de modo a demonstrar a eficácia do Apriori na geração de regras de
associação e estabelecer padrões de preferências de POI entre grupos turísticos.
Os resultados dos testes mostram que o algoritmo Apriori é capaz de incentivar uma
abordagem mais personalizada de POI para incluir ou excluir da recomendação, através dos
padrões que encontra entre as preferências e características dos turistas, melhorando assim as
sugestões personalizadas, conscientes do contexto em que se enquadram, apesar de conferir
limitações no que toca à estabilidade do algoritmo em grandes conjuntos de dados. Assim, estas
descobertas mostram um desempenho confiável na criação de listas de pontos de interesse (POI)
para incluir ou excluir das recomendações, e definem o caminho para futuras melhorias nas
tecnologias de sistemas de recomendação.
Recommendation systems (RS) are used in various sectors to predict, filter, or provide recommendations to users based on previous behaviors, patterns, or preferences. However, when it is necessary to make these predictions for groups, where multiple users with different preferences are involved, traditional RS face more difficulty in providing recommendations that meet their needs. Group recommendation systems (GRS) aim to solve this problem by offering personalized recommendations, usually based on the aggregated preferences of each group member. The tourism sector is an example of an area where this type of recommendation is complex. To improve suggestions given to tourist groups, this project proposed integrating association rules, through the Apriori algorithm, into the current Multi-Agent Microservice (MAMS) of a tourism GRS prototype, Grouplanner. The Apriori algorithm is an important tool for mining association rules to identify patterns and associations between objects in a dataset. It can be used to discover associations between frequently visited points of interest (POI) by examining the travel preferences and experiences of group members. The main goals of the project were to improve POI selection, encourage a more personalized approach to recommendations, and dynamically modify recommendations according to individual group members' preferences. This was achieved through predictions resulting from the Apriori algorithm, integrating it into one of the microservices developed in .NET and hosted on Azure, MAMS, which is part of the GRS prototype. A customized Apriori algorithm was developed in C#, as well as the generation of association rules in line with the multi-agent system. Subsequently, tests were conducted to demonstrate the effectiveness of Apriori in generating association rules and establishing POI preference patterns among tourist groups. The results show that the Apriori algorithm can encourage a more personalized approach to which POI to include or exclude from recommendations, based on the patterns it identifies among tourists' preferences and characteristics. This improves personalized suggestions, aware of the context in which they are framed, although the algorithm presents limitations regarding stability with large datasets. These findings show reliable performance in creating lists of points of interest (POI) to include or exclude from recommendations and pave the way for future improvements in recommendation system technologies.
Recommendation systems (RS) are used in various sectors to predict, filter, or provide recommendations to users based on previous behaviors, patterns, or preferences. However, when it is necessary to make these predictions for groups, where multiple users with different preferences are involved, traditional RS face more difficulty in providing recommendations that meet their needs. Group recommendation systems (GRS) aim to solve this problem by offering personalized recommendations, usually based on the aggregated preferences of each group member. The tourism sector is an example of an area where this type of recommendation is complex. To improve suggestions given to tourist groups, this project proposed integrating association rules, through the Apriori algorithm, into the current Multi-Agent Microservice (MAMS) of a tourism GRS prototype, Grouplanner. The Apriori algorithm is an important tool for mining association rules to identify patterns and associations between objects in a dataset. It can be used to discover associations between frequently visited points of interest (POI) by examining the travel preferences and experiences of group members. The main goals of the project were to improve POI selection, encourage a more personalized approach to recommendations, and dynamically modify recommendations according to individual group members' preferences. This was achieved through predictions resulting from the Apriori algorithm, integrating it into one of the microservices developed in .NET and hosted on Azure, MAMS, which is part of the GRS prototype. A customized Apriori algorithm was developed in C#, as well as the generation of association rules in line with the multi-agent system. Subsequently, tests were conducted to demonstrate the effectiveness of Apriori in generating association rules and establishing POI preference patterns among tourist groups. The results show that the Apriori algorithm can encourage a more personalized approach to which POI to include or exclude from recommendations, based on the patterns it identifies among tourists' preferences and characteristics. This improves personalized suggestions, aware of the context in which they are framed, although the algorithm presents limitations regarding stability with large datasets. These findings show reliable performance in creating lists of points of interest (POI) to include or exclude from recommendations and pave the way for future improvements in recommendation system technologies.
Description
Financiado
Keywords
Apriori algorithm Association rules Group recommendation systems Multi-agent systems Multi-agent microservices Algoritmo Apriori, Regras de associação Microserviços multiagente Sistemas de recomendação para grupos Sistemas multiagente