Banca de QUALIFICAÇÃO: FREDERICO JOSE DE ATHAYDE GUIMARAES

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : FREDERICO JOSE DE ATHAYDE GUIMARAES
DATE: 27/08/2025
TIME: 13:30
LOCAL: Sala do PPGER
TITLE:

CHALLENGES OF APPLYING ARTIFICIAL INTELLIGENCE TOOLS FOR ECONOMIC ANALYSIS: THE EVOLUTION OF MACHINE LEARNING ALGORITHMS AND NEURAL NETWORKS FOR APPLICATION IN REGIONAL ECONOMY STUDIES

KEY WORDS:

Artificial Neural Networks; Machine Learning; Economic Analysis; Algorithmic Modeling; Artificial Intelligence; Natural Language Processing; Regional Economics; Interior Development RJ.

PAGES: 52
BIG AREA: Ciências Sociais Aplicadas
AREA: Economia
SUMMARY:

The development of Artificial Intelligence mechanisms has been heavily supported by Artificial Neural Network (ANN) models and Machine Learning (ML) algorithms. ML algorithms use data to obtain experimental knowledge (learning) and generate generalist predictive models that can be applied to data outside the sample space. ANNs are defined as massively parallel processors that simulate the human brain with the intention of collecting empirical evidence and using it in experimental knowledge processing (ML) by simulating the interactions between neural networks. This development has driven the improvement of stochastic data analysis and expanded the universe of statistical modeling to include the use of algorithms (algorithmic modeling) that allow the definition of interdependence relationships even in scenarios where the data mechanism is unknown. Most economic modeling still relies almost exclusively on stochastic data models supported by theories based on linear relationships of predictive variables (Breiman 2001). This approach has limitations when it comes to a wide range of interesting current problems. Algorithmic modeling, both theoretically and practically, has developed rapidly. It can be used both on large, complex data sets and as a more accurate and informative alternative to traditional modeling on smaller data sets. If we aim to use data to solve critical problems, and if we consider econometrics essentially as a tool for decision-making under uncertainty (Chamberlain, 2000), we need to understand algorithmic tools, enabling us to add alternatives to more traditional models and enhance analytical possibilities by adopting a more diverse set of tools (Athey et al. 2019). Furthermore, the use of these tools combined with natural language processing (NLP) for the extraction and analysis of text data enables an innovative approach that can contribute not only to a more precise understanding of the specific conditions of economic analysis but also, based on methodological development, serve as a model for the use of these tools in future regional economic analyses. Given their unique characteristics, these characteristics are not defined in generic macroeconomic models nor standardized in official structured data. Therefore, it is necessary to extract and process information from unstructured data (free text) present in social media interactions and regional news sites. In this work, we hope to demonstrate the guiding principles of algorithmic modeling, the benefits and limitations for economic analysis compared to traditional stochastic methods, and, secondarily, to apply these tools in an empirical comparative analysis of agricultural development in the interior of Rio de Janeiro state compared to that of the interior of São Paulo state.

COMMITTEE MEMBERS:
Externo à Instituição - EDSON ZAMBON MONTE - UFES
Presidente - ***.915.177-** - EVERLAM ELIAS MONTIBELER - UFES
Interno - 2639882 - FELIPE LEITE COELHO DA SILVA