The analysis of the huge amount of financial data, made available by electronic markets, calls for new models and techniques to effectively extract knowledge to be exploited in an informed decision-making process. The purpose of this thesis is to discuss how graphical models can be used to reason and perform actions in this context. In the first part of this thesis, I present a framework in which the concepts underlying Bayesian networks are discussed, also highlighting the contributions of those, among the most important Judea Pearl, who have successfully used them to support problem solving in a wide range of disciplines: computer science, engineering, medicine and, more recently, as a decision-making tool to improve risk analysis in the energy sector. In the second part of the thesis, I show how it is possible to use this type of models to address real financial problems, also resorting to sensitivity analysis, which is a technique that can help validate the probability parameters of a Bayesian network. In particular, we focus on the period from 01/10/2011 to 01/01/2021, using quarterly data referring to the main reference indices of the American economy: GDP of the Gross Domestic Product, Producer Price Index (PPI), Price Index to consumption (CPI, unemployment rate and WTI and Brent oil prices. The aim is precisely to see how the trend of these indicators have changed before and after the advent of COVID-19, together with the trend in the share price (and therefore also of the multiples P/E ratio and P/B ratio) of two companies operating in the Oils / Energy sector: Chevron Corporation and Devon Energy, relating to the Oil & Gas - International Integrated and Oil & Gas - US Exploration and Production industries respectively.
L'analisi dell'enorme mole di dati finanziari, resa disponibile dai mercati elettronici, richiede nuovi modelli e tecniche per estrarre efficacemente conoscenza da sfruttare in un processo decisionale informato. Lo scopo di questa tesi è discutere come i modelli grafici possono essere utilizzati per ragionare ed eseguire azioni in questo contesto. Nella prima parte di questa tesi, presento un framework in cui vengono discussi i concetti alla base delle reti bayesiane, evidenziando anche i contributi di coloro, tra i più importanti Judea Pearl, che le hanno utilizzate con successo per risoluzione di problemi in una vasta gamma di discipline: informatica, ingegneria, medicina e, più recentemente, come strumento decisionale per migliorare l'analisi dei rischi nel settore energetico. Nella seconda parte della tesi, mostro come sia possibile utilizzare questo tipo di modelli per affrontare problemi finanziari reali, ricorrendo anche all'analisi di sensitività, che è una tecnica che può aiutare a validare i parametri di probabilità di una rete bayesiana. In particolare, ci concentriamo sul periodo dal 01/10/2011 al 01/01/2021, utilizzando dati trimestrali riferiti ai principali indici di riferimento dell'economia americana: PIL del Prodotto Interno Lordo, Indice dei prezzi alla produzione (PPI), Indice dei prezzi al consumo (CPI, tasso di disoccupazione e prezzi del petrolio WTI e Brent. L'obiettivo è proprio quello di vedere come è cambiato l'andamento di questi indicatori prima e dopo l'avvento del COVID-19, insieme all'andamento del prezzo delle azioni (e quindi anche dei multipli P/E ratio e P/B ratio) di due società operanti nel settore Oils/Energy: Chevron Corporation e Devon Energy, relative rispettivamente ai settori Oil & Gas - International Integrated e Oil & Gas - US Exploration e Production.
Bayesian Networks: a decision tool for the energy market
FERRARESE, SAMUELE
2020/2021
Abstract
The analysis of the huge amount of financial data, made available by electronic markets, calls for new models and techniques to effectively extract knowledge to be exploited in an informed decision-making process. The purpose of this thesis is to discuss how graphical models can be used to reason and perform actions in this context. In the first part of this thesis, I present a framework in which the concepts underlying Bayesian networks are discussed, also highlighting the contributions of those, among the most important Judea Pearl, who have successfully used them to support problem solving in a wide range of disciplines: computer science, engineering, medicine and, more recently, as a decision-making tool to improve risk analysis in the energy sector. In the second part of the thesis, I show how it is possible to use this type of models to address real financial problems, also resorting to sensitivity analysis, which is a technique that can help validate the probability parameters of a Bayesian network. In particular, we focus on the period from 01/10/2011 to 01/01/2021, using quarterly data referring to the main reference indices of the American economy: GDP of the Gross Domestic Product, Producer Price Index (PPI), Price Index to consumption (CPI, unemployment rate and WTI and Brent oil prices. The aim is precisely to see how the trend of these indicators have changed before and after the advent of COVID-19, together with the trend in the share price (and therefore also of the multiples P/E ratio and P/B ratio) of two companies operating in the Oils / Energy sector: Chevron Corporation and Devon Energy, relating to the Oil & Gas - International Integrated and Oil & Gas - US Exploration and Production industries respectively.È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per contatti: unitesi@unipv.it
https://hdl.handle.net/20.500.14239/1379