Ο σκοπός της παρούσας διπλωματικής εργασίας είναι η δημιουργία μιας μεθόδου για την εξαγωγή αποδοτικών και ερμηνεύσιμων επενδυτικών κανόνων με χρήση μεθόδων υπολογιστικής νοημοσύνης. Οι επενδυτικοί αυτοί κανόνες εξάγονται αυτόματα από το σύστημα και υποδεικνύουν τη στρατηγική που πρέπει να ακολουθήσει ένας χρήστης. Αποκαλύπτουν το συσχετισμό των εισόδων και παρέχουν πληροφορίες για κερδοφόρες επενδυτικές στρατηγικές. Η υπολογιστική νοημοσύνη (computational intelligence) αποτελεί παρακλάδι της τεχνητής νοημοσύνης το οποίο περιλαμβάνει τον σχεδιασμό και την ανάπτυξη θεωριών και μεθόδων, βασιζόμενη στην κατανόηση της βιολογίας και της προσπάθειας για εφαρμογή σε προβλήματα του πραγματικού κόσμου. Ένα σύστημα είναι υπολογιστικά ευφυές όταν: ασχολείται μόνο με αριθμητικά (χαμηλού επιπέδου) δεδομένα, έχει συστατικά αναγνώρισης προτύπων, δεν χρησιμοποιεί γνώσεις στην μορφή της τεχνητή νοημοσύνης και επιπλέον, εμφανίζει i) υπολογιστική προσαρμοστικότητα, ii) υπολογιστική ανοχή σε σφάλματα, iii) επιτάχυνση που προσεγγίζει την ανθρώπινη, και iv) τα ποσοστά σφάλματός του προσεγγίζουν την ανθρώπινη απόδοση. Οι αλγόριθμοι της υπολογιστικής νοημοσύνης αποτελούνται από μοντέλα που εκπαιδεύονται από τα παραδείγματα με την βοήθεια ενός δασκάλου (επιβλεπόμενη μάθηση) και μοντέλα τα οποία προσαρμόζονται μόνα τους (μη επιβλεπόμενη μάθηση). Το πρόβλημα στις παρούσες προσεγγίσεις για την πρόβλεψη οικονομικών δεικτών εντοπίζεται στην μη ερμηνευσιμότητα των αποτελεσμάτων. Ενώ υπάρχουν δυνατά υπολογιστικά μοντέλα, όπως οι γενετικοί αλγόριθμοι και οι μηχανές διανυσμάτων υποστήριξης, τα αποτελέσματα τους δεν είναι ερμηνεύσιμα. Από την άλλη τα μοντέλα της ασαφούς λογικής ενώ παρουσιάζουν ερμηνεύσιμα αποτελέσματα δεν έχουν την δύναμη να παράγουν αποδοτικούς κανόνες. Το μοντέλο που προτείνεται σε αυτή την εργασία συνδυάζει τις τρεις προαναφερθείσες μεθόδους ονομάζεται ESVM-Fuzzy Inference Trader. Το προτεινόμενο μοντέλο χρησιμοποιείται για την πρόβλεψη των δεικτών DAX και FTSE 100. Τα αποτελέσματα του ESVM Fuzzy Inference Trader ξεπέρασαν σε απόδοση τις παραδοσιακές μεθόδους καθώς και μια εξελιγμένη τεχνική μηχανικής μάθησης. / The purpose of the present thesis is to develop a method for extracting efficient and interpretable investment rules, using methods of Computational Intelligence. The investment rules are automatically extracted from the system and suggest the strategy to be followed by a user. They are revealing the correlation between inputs and provide information on profitable investment strategies. Computational intelligence (CI) constitutes a subbranch of Artificial Intelligence (AI) that includes the design and development of theories and methods with a sound biological understanding alongside their application to solve real world problems. A system is computationally intelligent when it deals with only numerical (low level) data, has pattern recognition components, does not use knowledge in the AI sense and additionally when it (begins to) exhibit i) computational adaptivity, ii) computational fault tolerance, iii) speed approaching human-like turn around and iv) error rates that approximate human performance. The CI algorithms consist of models that are trained from examples with the aid of a tutor (supervised learning) and models that are self-adapted (unsupervised learning) The problem in the current approaches for predicting economic indicators is the non-interpretability of results. While there are strong computational models, such as genetic algorithms and support vector machines their results are not interpretable. On the other hand fuzzy logic models create interpretable results, but lack the power to produce efficient rules. The model proposed in this paper combines the three previous methods is called ESVM-Fuzzy Inference Trader. The proposed model is used to predict the indices DAX and FTSE 100. The results of ESVM Fuzzy Inference Trader outperformed traditional methods as well as an advanced machine learning technique.

Submitted by Navjeet Singh Gill (navjeetgill7@gmail.com) on 2018-02-23T18:00:11Z No. of bitstreams: 1 2017-18_Navjeet_Gill v6.pdf: 1539898 bytes, checksum: c559bf4e5d8539b1b89e7d6f75555716 (MD5) / Approved for entry into archive by Josineide da Silva Santos Locatelli (josineide.locatelli@fgv.br) on 2018-02-23T18:06:56Z (GMT) No. of bitstreams: 1 2017-18_Navjeet_Gill v6.pdf: 1539898 bytes, checksum: c559bf4e5d8539b1b89e7d6f75555716 (MD5) / Made available in DSpace on 2018-02-26T12:51:55Z (GMT). No. of bitstreams: 1 2017-18_Navjeet_Gill v6.pdf: 1539898 bytes, checksum: c559bf4e5d8539b1b89e7d6f75555716 (MD5) Previous issue date: 2018-01-16 / This empirical research examines the relationship between board’s diversity and firm performance, providing a comprehensive quantitative analysis between diversity factors (demographic - gender and race, cognitive - age, education, role and network) and financial factors (ROA, ROE, ROIC, asset turnover and current ratio) in the component companies of the FTSE 100 index. The dataset also includes a wide array of information about 1053 board members. The results indicate that a diverse board positively impacts ROA, ROIC, asset turnover and current ratio but were insignificant for ROE. It proves that diversity leads to better social reputation, performance and financial performance. / Esta pesquisa empírica examina a relação entre a diversidade do conselho e o desempenho da empresa, fornecendo uma análise quantitativa significativa entre os fatores de diversidade (demográfico - gênero e raça, idade cognitiva, educação, papel e rede) e fatores financeiros (ROA, ROE, ROIC, rotatividade de ativos e liquidez geral) nas empresas que compõem o índice FTSE 100. O conjunto de dados também inclui uma ampla gama de informações sobre 1053 membros do conselho. Os resultados indicam que um conselho diversificado impacta positivamente no ROA, ROIC, rotatividade de ativos e liquidez geral, mas são insignificantes para ROE. Isto demonstra que a diversidade leva a uma melhor reputação social, performance e desempenho financeiro.

The availability of ultra high-frequency (UHF) data on transactions has revolutionised data processing and statistical modelling techniques in finance. The unique characteristics of such data, e.g. discrete structure of price change, unequally spaced time intervals and multiple transactions have introduced new theoretical and computational challenges. In this study, we develop a Bayesian framework for modelling integer-valued variables to capture the fundamental properties of price change. We propose the application of the zero inflated Poisson difference (ZPD) distribution for modelling UHF data and assess the effect of covariates on the behaviour of price change. For this purpose, we present two modelling schemes; the first one is based on the analysis of the data after the market closes for the day and is referred to as off-line data processing. In this case, the Bayesian interpretation and analysis are undertaken using Markov chain Monte Carlo methods. The second modelling scheme introduces the dynamic ZPD model which is implemented through Sequential Monte Carlo methods (also known as particle filters). This procedure enables us to update our inference from data as new transactions take place and is known as online data processing. We apply our models to a set of FTSE100 index changes. Based on the probability integral transform, modified for the case of integer-valued random variables, we show that our models are capable of explaining well the observed distribution of price change. We then apply the deviance information criterion and introduce its sequential version for the purpose of model comparison for off-line and online modelling, respectively. Moreover, in order to add more flexibility to the tails of the ZPD distribution, we introduce the zero inflated generalised Poisson difference distribution and outline its possible application for modelling UHF data.