Use this link to cite:
http://hdl.handle.net/2183/31265 Bandidos multibrazo para selección de cartera
Loading...
Identifiers
Publication date
Authors
Vega Rodríguez, Daniel Sergio
Advisors
Other responsabilities
Enxeñaría informática, Grao en
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen] Este proyecto consiste en la implementación y validación de modelos para recomendación
de cartera (portfolio recommendation). Para ello será necesario el estudio, diseño, desarrollo e
implementación de un software capaz de asistir al usuario en la selección de valores bursátiles
en los que invertir. Nos centraremos en un los modelos denominados como “Bandidos Bayesianos”
para poder estimar qué valores ofrecen mejor rentabilidad al usuario combinando explotación
y exploración. Dichos modelos seleccionarán valores en función de una estimación
de probabilidades construida a partir de conjuntos de datos que describen el comportamiento
de dichos valores en el pasado. Los valores más prometedores serán usualmente recomendados
al usuario, con la esperanza de obtener un resultado positivo en la inversión futura. Las
recomendaciones dadas por el algoritmo tenderán a variar a medida que nuevos datos son
integrados en el modelo, ya que estos usualmente revelarán cambios en el comportamiento
de los valores observados.
[Abstract] This project revolves around the implementation and validation of models meant for portfolio recommendation. To accomplish this, the investigation, design, development and implementation of a demonstrative application will be necessary. The resulting tool will assist the user in making decisions over stock values. This project will focus on a series of models based on the Bayesian Bandit agent to estimate which values are likely to be promising. Said models will choose values based on a collection of likelihood random variables which are constructed from datasets that describe the historic behaviour of the considered values. The most promising values will be usually recommended to the user. That is, hoping to obtain a positive result in the future investment. Recommendations given by the algorithm will tend to change over new data additions to the model. Those additions will often reveal changes in the behaviour of the observed stock values.
[Abstract] This project revolves around the implementation and validation of models meant for portfolio recommendation. To accomplish this, the investigation, design, development and implementation of a demonstrative application will be necessary. The resulting tool will assist the user in making decisions over stock values. This project will focus on a series of models based on the Bayesian Bandit agent to estimate which values are likely to be promising. Said models will choose values based on a collection of likelihood random variables which are constructed from datasets that describe the historic behaviour of the considered values. The most promising values will be usually recommended to the user. That is, hoping to obtain a positive result in the future investment. Recommendations given by the algorithm will tend to change over new data additions to the model. Those additions will often reveal changes in the behaviour of the observed stock values.
Description
Editor version
Rights
Atribución-NoComercial-CompartirIgual 3.0 España







