Geri Dön

Complex mutual information-theoretic stock networks

Başlık çevirisi mevcut değil.

  1. Tez No: 719642
  2. Yazar: SERKAN ALKAN
  3. Danışmanlar: DR. KHALDOUN KHASHANAH
  4. Tez Türü: Doktora
  5. Konular: İstatistik, Statistics
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2019
  8. Dil: İngilizce
  9. Üniversite: Stevens Institute of Technology
  10. Enstitü: Yurtdışı Enstitü
  11. Ana Bilim Dalı: Belirtilmemiş.
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 133

Özet

Özet yok.

Özet (Çeviri)

Financial markets can be characterized as complex systems because of interactions between heterogeneous components and existing nonlinearities. Network theory can be used to model the financial market in which nodes can be stocks, commodities or currencies. The current state of the art is to measure distances between nodes using Pearson correlations. In this thesis, mutual information (MI) is applied to describe similarities between assets. MI is a general measure of dependency and can detect linear and non-linear relationships whereas Pearson correlation fails to detect the nonlinear relationships. This dissertation consists of three essays. In essay 1, we propose a comprehensive comparison approach between the mutual information (MI) metric and the Pearson correlation metric. We find that networks constructed by these two measures become very different during depending on market regimes and network topological structure. We pay special attention to the cases where two measures are strongly mismatched and examine the reasons. Relationship between the entropy and the moments of daily stock return distributions are investigated and we find very strong relationship between entropy and the kurtosis. We compare the performance of MI and correlation techniques in terms of identification of coherent and well-separated stock communities and results show that MI approaches perform better in crisis and non-crisis periods. In essay 2, we analyzed how local, mesoscopic and global topological properties of mutual information based stock networks evolve annually. To detect and quantify the impact of a major crisis on the market network structure, we propose to use iv the information-theoretic quantifiers. We observe that the entropy of the system relatively reduces during crisis regimes. We propose to use classic internal cluster indices and information-theoretic quantifiers to capture the topological evolution of sectors by treating them as fixed communities, i.e. labels of stocks do not change over time in the system. We find that the structural changes of the financial sector during the subprime crisis and it has the strongest correlation with markets in terms of structural uncertainty. In essay 3, we analyze how the homogeneity in each aggregation level of Global Industry Classification System (GICS) scheme changes over time and identify the industries which have more homogeneous structure than others in terms of stock returns comovement. We propose a mesoscopic approach in terms of network-theoretic framework to describe the relationship between industry groups. We investigate the time evolution of the interaction structure of the industry groups and identify the important ones in the market. Our analysis reveals that during a crisis period, banks, as an industry group, become more important in the market and the local interaction structure of industry groups becomes much simpler or a star network topology. In conclusion, we find that some asset pairs have non-linear relationships and MI provides a better alternative measure to define the links in financial networks. MI network and Pearson network become very different at local and global topological scales during various regimes. Community detection algorithms yield well-separated stock communities with MI networks comparing Pearson ones. Besides the classic network measures, information-theoretic measures provide an information advantage in revealing nonlinear dependencies to quantify and detect financial crises and regime switching.

Benzer Tezler

  1. Portföy yönetiminde dinamik varlık yönetim stratejileri

    Dynamic asset allocation strategies in portfolio management

    MUSTAFA DUMAN

    Yüksek Lisans

    Türkçe

    Türkçe

    2000

    BankacılıkMarmara Üniversitesi

    Sermaye Piyasası ve Borsa Ana Bilim Dalı

    YRD. DOÇ. DR. ÖZLEM KOÇ

  2. Uluslararası fon piyasaları ve döviz kredileri mekanizması (analitik bir yaklaşım)

    A Short history of the foreign exchange markets

    ADNAN YİĞİT

    Yüksek Lisans

    Türkçe

    Türkçe

    1994

    BankacılıkMarmara Üniversitesi

    Uluslararası Bankacılık ve Finans Ana Bilim Dalı

    PROF. DR. İLHAN ULUDAĞ

  3. Topics in signal processing: Applications in genomics and genetics

    Başlık çevirisi yok

    ABDÜLKADİR ELMAS

    Doktora

    İngilizce

    İngilizce

    2016

    GenetikColumbia University

    Prof. XIAODONG WANG

  4. Türkçe fonemler için en uygun ana dalgacık fonksiyonunun araştırılması

    The investigation of optimum mother wavelet function for turkish phonemes

    ÖZKAN ARSLAN

    Yüksek Lisans

    Türkçe

    Türkçe

    2014

    Elektrik ve Elektronik MühendisliğiEge Üniversitesi

    Elektrik-Elektronik Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. ERKAN ZEKİ ENGİN

  5. İlköğretim matematik öğretmenleri ile akademisyenlerin yüz-yüze ve web-tabanlı ortamda bilgi paylaşımlarının değerlendirilmesi

    The evaluation of information sharing processes of primary mathematics teachers and academicians in face to face and web-based contexts

    FATİH BAŞ

    Doktora

    Türkçe

    Türkçe

    2013

    Eğitim ve ÖğretimAtatürk Üniversitesi

    İlköğretim Ana Bilim Dalı

    PROF. AHMET IŞIK