Yüksek frekanslı mali piyasa verileri üzerinde yapay sinir ağı tabanlı karar verici tahmin modelleri

Loading...
Thumbnail Image

Date

2017

Journal Title

Journal ISSN

Volume Title

Publisher

Research Projects

Organizational Units

Organizational Unit
Industrial Engineering
(1998)
Industrial Engineering is a field of engineering that develops and applies methods and techniques to design, implement, develop and improve systems comprising of humans, materials, machines, energy and funding. Our department was founded in 1998, and since then, has graduated hundreds of individuals who may compete nationally and internationally into professional life. Accredited by MÜDEK in 2014, our student-centered education continues. In addition to acquiring the knowledge necessary for every Industrial engineer, our students are able to gain professional experience in their desired fields of expertise with a wide array of elective courses, such as E-commerce and ERP, Reliability, Tabulation, or Industrial Engineering Applications in the Energy Sector. With dissertation projects fictionalized on solving real problems at real companies, our students gain experience in the sector, and a wide network of contacts. Our education is supported with ERASMUS programs. With the scientific studies of our competent academic staff published in internationally-renowned magazines, our department ranks with the bests among other universities. IESC, one of the most active student networks at our university, continues to organize extensive, and productive events every year.

Journal Issue

Abstract

Yüksek frekanslı mali piyasa verilerinin tahmin edilmeleri ve modellenmeleri, tamamen imkansız olmasa da, stokastik prosesler ve çok sayıda rastgele faktör araya girdiğinden, çok zordur. Bu çalışmada; mali zaman serisi tahmini ve karar alınmasına yardımcı, yeni bir Yapay Zeka modeli tasarlanmış ve geliştirilmiştir. Döviz çiftlerinin tahmin başarı yüzdelerinin artırılması iki şekilde irdelenmiştir: ilk olarak mühendislik tasarımlarıyla tahmin modellerinin ortaya çıkmasını sağlayacak metodolojinin kullanımı ile, kaos teorisi ve fiyat grafiklerinin görsel özelliklerinin devreye sokulması, ikinci olarak da en güncel ve güçlü yöntemlerin kullanılması. Görsel özellikleri kullanan yaklaşımda, aynı Yapay Görmedeki gibi, yoğunluk bölgeleri ile birlikte şekli tanımlayan yüksek ve düşük fiyat değerlerinden faydalanılmaktadır. Modellemede esas olarak Yapay Sinir Ağları uygulanmakta olup; diğer popüler yöntemlerden, İleri Gradyant Artırımı ve Destek Vektör Makinesi de karşılaştırma amaçlı olarak kullanılmıştır. Tasarlanan system ayrıca yazılımla gerçek zamanlı alım-satım robotu olarak da kodlanmıştır. Testler ve simulasyonlarda tatmin edici sonuçlar elde edilmiştir.
High frequency financial data are somewhat hard to model or predict, if not totally impossible, as stochastic processes and many other random factors are involved. In this thesis; a novel Artificial Intelligence model is designed and developed for financial time series prediction and decision making. Possibility to enhance prediction accuracy for foreign exchange rates is investigated in two ways: first applying an outside the box approach by bringing about methodology and techniques to facilitate the use of predictive models in engineering design to model price graphs by exploiting their visual properties together with principles of chaos theory, and secondly employing the most efficient methods to detect patterns to classify the direction of movement. The approach that exploits the visual properties of price graphs makes use of density regions along with high and low values describing the shape just as in Machine Vision. Mainly Artificial Neural Networks are used in modeling. However, other state-of-the-art methods; Extreme Gradient Boosting and Support Vector Machine are too used for comparison. The designed system is also software coded as a real-time trading robot. Comparable prediction results and profits are achieved in tests and simulations.

Description

Keywords

Endüstri ve Endüstri Mühendisliği, Industrial and Industrial Engineering

Turkish CoHE Thesis Center URL

Citation

WoS Q

Scopus Q

Source

Volume

Issue

Start Page

0

End Page

104