| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 858.19 KB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Esta tese tem como objetivo investigar e desenvolver sistemas de recomendação de entidades biomédicas
relacionadas ao autismo, com o propósito de fornecer ferramentas eficazes para a descoberta de informações personalizadas e relevantes a investigadores e profissionais de saúde. O estudo explora três abordagens distintas de divisão de dados e avalia seu impacto no desempenho do sistema de recomendação.
Foram recolhidos e processados dados biomédicos relevantes, incluindo informações sobre doenças ancestrais do autismo, artigos do PubMed e detalhes dos autores. De forma a testar como é que os dados
se comportam num sistema de recomendação, foi desenvolvido um programa em python utilizando a
biblioteca surprise. Resultados de métricas de avaliação revelam detalhes valiosos sobre o desempenho
do sistema em diferentes cenários. Este trabalho contribui para a compreensão dos desafios na recomendação de entidades biomédicas relacionadas ao autismo e destaca a importância das estratégias de divisão
de dados. Futuras pesquisas são incentivadas a explorar técnicas e estratégias adicionais para melhorar a
precisão e relevância das recomendações em contextos complexos.
This thesis aims to investigate and develop Recommender systems for recommending biomedical entities related to autism, with the purpose of providing effective tools for personalized and relevant information discovery to researchers and healthcare professionals. The study explores three distinct data splitting approaches and evaluates their impact on the recommendation system’s performance. Relevant biomedical data, including information on autism-related ancestral diseases, PubMed articles, and author details, were collected and processed. A Python program, using the Surprise library, was developed as the foundation for the recommendation system. Evaluation metric results reveal valuable insights into the system’s performance in different scenarios. This work contributes to the understanding of challenges in recommending biomedical entities related to autism and underscores the importance of data splitting strategies. Future research is encouraged to explore additional techniques and strategies to enhance the accuracy and relevance of recommendations in complex contexts.
This thesis aims to investigate and develop Recommender systems for recommending biomedical entities related to autism, with the purpose of providing effective tools for personalized and relevant information discovery to researchers and healthcare professionals. The study explores three distinct data splitting approaches and evaluates their impact on the recommendation system’s performance. Relevant biomedical data, including information on autism-related ancestral diseases, PubMed articles, and author details, were collected and processed. A Python program, using the Surprise library, was developed as the foundation for the recommendation system. Evaluation metric results reveal valuable insights into the system’s performance in different scenarios. This work contributes to the understanding of challenges in recommending biomedical entities related to autism and underscores the importance of data splitting strategies. Future research is encouraged to explore additional techniques and strategies to enhance the accuracy and relevance of recommendations in complex contexts.
Description
Tese de mestrado, Bioinformática e Biologia Computacional, 2023, Universidade de Lisboa, Faculdade de Ciências
Keywords
Sistemas de Recomendação Autismo Surprise Biomédica PubMed Teses de mestrado - 2024