Una comprensiva revisión de los métodos de recomendación basados en técnicas probabilísticas
Loading...
Publication date
Advisors
Department
Research group
Center
Abstract
Esta investigación tiene como objetivo utilizar un método de recomendación hibrido
basado en técnicas probabilísticas y de modelado de tópicos que brinde al usuario recomendaciones
más ajustadas frente a los modelos de recomendación tradicionales. Este artículo presenta una
revisión comprensiva de los métodos de recomendación para sistemas basado en contenido y filtrado
colaborativo. Entre los métodos analizados están las Matrices de Factorización Probabilística y el
método de Asignación Latente de Dirichlet. La revisión de la literatura entorno a estos modelos se
centra en la identificación de problemas y cuestiones abiertas que pueden ser abarcadas para futuras
investigaciones. Se analiza el funcionamiento de algunos modelos de recomendación que integran
técnicas de factores latentes y de modelado de tópicos, que serán de base para comparar los
resultados obtenidos con el modelo híbrido.
This research aims to use a hybrid recommendation method based on probabilistic techniques and topics modeling that provide recommendations most close fitting the user compared to other traditional recommendation models. We carry out a comprehensive review of the recommended methods for content-based systems and collaborative filtering, mainly in the domain of recommending movies. The methods discussed are the matrix factorization and Latent Dirichlet Allocation method. The literature review around these models focuses on identifying problems and open issues that may be covered for future researches. Also, we analyzed the recommendation models that integrant latent factor methods and topics modeling, which will be used to compare results obtained with the hybrid model
This research aims to use a hybrid recommendation method based on probabilistic techniques and topics modeling that provide recommendations most close fitting the user compared to other traditional recommendation models. We carry out a comprehensive review of the recommended methods for content-based systems and collaborative filtering, mainly in the domain of recommending movies. The methods discussed are the matrix factorization and Latent Dirichlet Allocation method. The literature review around these models focuses on identifying problems and open issues that may be covered for future researches. Also, we analyzed the recommendation models that integrant latent factor methods and topics modeling, which will be used to compare results obtained with the hybrid model







