Üç Boyutlu Evrişimsel Sinir Ağları İle Finansal Zaman Serisi Analizi
No Thumbnail Available
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Open Access Color
OpenAIRE Downloads
OpenAIRE Views
Abstract
Yapay sinir ağları, finansal zaman serisi tahmininde son yıllarda oldukça yaygın biçimde kullanılmaktadır. Özellikle karmaşık, doğrusal olmayan ilişkileri modelleyebilme yetenekleri sayesinde geleneksel yöntemlerin önüne geçmektedir. Bu kapsamda zaman bağımlılıklarını etkili bir şekilde öğrenebilen Kısa Uzun Süreli Hafıza (LSTM) tabanlı modeller finansal verinin tahmininde oldukça fazla kullanılmaktadır. Evrişimsel Sinir Ağları (CNN) da literatürde kendisine yer bulmaya başlamış ve çeşitli çalışmalarda başarılı sonuçlar elde edilmiştir. Bu çalışmada ise klasik CNN yaklaşımlarının ötesine geçilerek, konvolüsyon işlemi üç boyutlu veri yapısı üzerinde gerçekleştirilmiştir. Bu yeni yaklaşımda, veri hazırlama süreci detaylı şekilde tasarlanmıştır. İlk olarak, teknik analizde yaygın olarak kullanılan 20 farklı indikatör seçilmiş; her bir indikatör, 5 ile 25 gün arasındaki 20 farklı periyot için hesaplanarak, her gün için 20x20 boyutunda iki boyutlu bir matris oluşturulmuştur. Ardından, bu iki boyutlu veriye ardışık 20 günlük geçmiş veriler eklenmiş ve sonuç olarak her bir gün için 20x20x20 boyutunda üç boyutlu veri elde edilmiştir. Bu veri yapısı, CNN modelinin zaman, indikatör ve periyot bilgilerini aynı anda işlemleyerek daha karmaşık örüntüleri öğrenmesine olanak tanımaktadır. Etiketleme sürecinde ise her gün için fiyat hareketine göre 'Al', 'Sat' ve 'Tut' olmak üzere üç sınıftan biri atanmıştır. Model, bu sınıflandırmayı öğrenerek gelecekteki alım-satım sinyallerini tahmin etmeyi hedeflemiştir. Modelin tahminlerine dayanarak oluşturulan alım-satım stratejileri, 2022–2024 yılları arasında toplam 750 işlem gününü kapsayan bir dönem boyunca Dow Jones 30 endeksine dahil olan hisseler ile çeşitli borsa yatırım fonları (ETF) üzerinde test edilmiştir. Gerçekleştirilen işlemler sonucunda, hisseler için yıllık ortalama %23.87, ETF'ler için ise %19.40 oranında getiri sağlanmış ve bu sonuçlar geleneksel 'al ve tut' stratejisine kıyasla daha yüksek performans göstermiştir. Böylece önerilen modelin, zaman serisi verilerinde teknik göstergelerin çok boyutlu temsilinden faydalanarak etkili bir tahmin performansı sergileyebildiği ortaya konmuştur.
Artificial neural networks have become widely used in recent years for financial time series forecasting. Their ability to model complex, nonlinear relationships has allowed them to surpass traditional methods in many applications. Among these, LSTM-based models, which can effectively capture temporal dependencies, are extensively utilized for predicting financial data. Convolutional Neural Networks (CNNs) have also gained traction in the literature and have shown promising results in various studies. In this study, a novel CNN-based approach is proposed, extending beyond classical CNN implementations by performing convolution operations over a three-dimensional data structure. The data preprocessing phase was carefully designed. First, 20 commonly used technical analysis indicators were selected. Each indicator was calculated using 20 different time periods ranging from 5 to 25 days, resulting in a 20×20 two-dimensional matrix for each day. Subsequently, consecutive data from the past 20 days were added to this matrix, forming a three-dimensional input of size 20×20×20 for each day. This data structure enables the CNN model to simultaneously process information related to time, indicator types, and period lengths, thus allowing it to capture more complex patterns. In the labeling phase, each day was assigned one of three labels: 'Buy', 'Sell', or 'Hold', based on the direction of the price movement. The model was trained to learn this classification task and to predict future trading signals. Based on the model's outputs, a trading strategy was implemented and tested over 750 trading days between 2022 and 2024, using both Dow Jones 30 stocks and selected exchange-traded funds (ETFs). As a result of the simulated trades, the strategy achieved an average annual return of 23.87% for stocks and 19.40% for ETFs, outperforming the traditional buy- and-hold strategy. These findings demonstrate that the proposed model can effectively leverage the multidimensional representation of technical indicators in time series data to achieve robust predictive performance.
Artificial neural networks have become widely used in recent years for financial time series forecasting. Their ability to model complex, nonlinear relationships has allowed them to surpass traditional methods in many applications. Among these, LSTM-based models, which can effectively capture temporal dependencies, are extensively utilized for predicting financial data. Convolutional Neural Networks (CNNs) have also gained traction in the literature and have shown promising results in various studies. In this study, a novel CNN-based approach is proposed, extending beyond classical CNN implementations by performing convolution operations over a three-dimensional data structure. The data preprocessing phase was carefully designed. First, 20 commonly used technical analysis indicators were selected. Each indicator was calculated using 20 different time periods ranging from 5 to 25 days, resulting in a 20×20 two-dimensional matrix for each day. Subsequently, consecutive data from the past 20 days were added to this matrix, forming a three-dimensional input of size 20×20×20 for each day. This data structure enables the CNN model to simultaneously process information related to time, indicator types, and period lengths, thus allowing it to capture more complex patterns. In the labeling phase, each day was assigned one of three labels: 'Buy', 'Sell', or 'Hold', based on the direction of the price movement. The model was trained to learn this classification task and to predict future trading signals. Based on the model's outputs, a trading strategy was implemented and tested over 750 trading days between 2022 and 2024, using both Dow Jones 30 stocks and selected exchange-traded funds (ETFs). As a result of the simulated trades, the strategy achieved an average annual return of 23.87% for stocks and 19.40% for ETFs, outperforming the traditional buy- and-hold strategy. These findings demonstrate that the proposed model can effectively leverage the multidimensional representation of technical indicators in time series data to achieve robust predictive performance.
Description
Keywords
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Bilim ve Teknoloji, Computer Engineering and Computer Science and Control, Science and Technology
Turkish CoHE Thesis Center URL
Fields of Science
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
66
Google Scholar™
Sustainable Development Goals
3
GOOD HEALTH AND WELL-BEING

7
AFFORDABLE AND CLEAN ENERGY

10
REDUCED INEQUALITIES

13
CLIMATE ACTION

16
PEACE, JUSTICE AND STRONG INSTITUTIONS

