Análise de padrões comportamentais na entropia de mercado, comparando modelos de base linear e não linear

Abstract

In the first approach, the title of this investigation may appear as a contradiction. The concept of entropy, due to its close relationship with chaos, can be seen as not only the unpredictability of the market, but also the distortion of information, making it impossible to find patterns. This investigation considers that financial markets, and the real economy, are not simply chaotic systems but chaotic systems with a behavioral aspect that is more prominent in shorter timeframes. Throughout this work, all the concepts included in the title, as well as all the others examined, will provide a cohesive and valid perspective through a relevant bibliographical review. This dissertation reconciles economic theory on financial markets with the field of computational engineering, focusing on the predictability of artificial intelligence tools applied to the foreign exchange market, with the support of technical analysis. The scope of this study includes comparing the predictive capacity of linear models against non-linear models. Classical theories, associated with linear models, intend to embellish the relationships between factors in the real world, making them within reach of human understanding. However, with the strong criticism of classical theory, interest in non-linear modeling that achieves superior predictive power has emerged, as verified in the results of this study. Lastly, it is important to note that all algorithms used in this study were written by the author.

Date of Award	13 Jul 2023
Original language	Portuguese
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Paulo Alves (Supervisor)