Thesis PRONÓSTICO DE VOLATILIDAD DEL MERCADO ACCIONARIO CHILENO MEDIANTE UN SISTEMA DE ENSAMBLE BASADO EN MODELOS HÍBRIDOS ANN-EGARCH
Date
2017
Authors
TORRES GONZÁLEZ, CAMILA FRANCISCA
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
En la presente memoria se aborda la temática del desarrollo de modelos de pronósticos precisos para la predecciónde la volatilidad financiera.El propósito del estudio consiste en el diseño e implementación de un sistema de ensamble basado en modeloshíbridos ANN-EGARCH mediante técnicas de programación para pronosticar la volatilidad del mercado accionariochileno. Dicho sistema se compone de tres modelos ANN-EGARCH bajo distintos supuestos distributivos que captan lanormalidad, la simetría y el exceso de curtosis, y una red neuronal adicional que permite complementar los resultadosindividuales y otorgar un pronóstico final.Los resultados demuestran que los sistemas de ensamble son efectivos en la predicción de volatilidad del índicebursátil IPSA del mercado accionario chileno, mejorando el desempeño predictivo de modelos de pronóstico popularestales como el modelo GARCH y los mismos modelos híbridos ANN-EGARCH que lo componen. Estos resultados sonrobustos y consistentes para diferentes arquitecturas de las redes neuronales que componen el sistema de ensamble ydiferentes horizontes de pronóstico empleados.
This thesis studies the development of accurate forecast models for the prediction of financial volatility.The use of the study consists of the design and implementation of an assembly system based on hybrid ANNEGARCHmodels through programming techniques to forecast the volatility of the Chilean stock market. This systemconsists of three ANN-EGARCH models under dierent distributive assumptions that capture normality, symmetry andexcess of kurtosis, and an additional neural network that allows complementing the individual results and giving a finalforecast.The results show that the assembly systems are eective in predicting the volatility of the IPSA stock index ofthe Chilean stock market, improving the predictive performance of popular forecast models such as the GARCH modeland the same ANN-EGARCH hybrid models that composes it. These results are robust and consistent for dierentneural network architectures that make up the assembly system and dierent prediction horizons used.
This thesis studies the development of accurate forecast models for the prediction of financial volatility.The use of the study consists of the design and implementation of an assembly system based on hybrid ANNEGARCHmodels through programming techniques to forecast the volatility of the Chilean stock market. This systemconsists of three ANN-EGARCH models under dierent distributive assumptions that capture normality, symmetry andexcess of kurtosis, and an additional neural network that allows complementing the individual results and giving a finalforecast.The results show that the assembly systems are eective in predicting the volatility of the IPSA stock index ofthe Chilean stock market, improving the predictive performance of popular forecast models such as the GARCH modeland the same ANN-EGARCH hybrid models that composes it. These results are robust and consistent for dierentneural network architectures that make up the assembly system and dierent prediction horizons used.
Description
Catalogado desde la version PDF de la tesis.
Keywords
IPSA , MERCADO DE VALORES , MODELOS EGARCH , PRONOSTICO DE VOLATILIDAD , REDES NEURONALES ARTIFICIALES , SISTEMA DE ENSAMBLE
