Back to Search

Deep transformer-based asset price and direction prediction

Title	Deep transformer-based asset price and direction prediction
Author	Gezici, Abdul Haluk Batur
Publication Date:	2024
Subject	Deep learning (Machine learning), Cryptocurrencies, Bitcoin, Finance, Mathematical models, Artificial intelligence, Financial applications
Type	Document
Language	English
Digital	Yes
Manuscript	No
Library:	Özyeğin University
Record ID	9637baaa-9e29-46be-adb2-c4b9735262b7
Library Location	Department of Artificial Intelligence
Date	2024
Sample Text	The field of algorithmic trading, driven by deep learning methodologies, has garnered substantial attention in recent times. Within this domain, transformers, convolutional neural networks, and patch embedding-based techniques have emerged as popular choices within the computer vision community. Here, inspired by the latest cutting edge computer vision methodologies and the existing work showing the capability of image-like conversion for time-series datasets, we apply more advanced transformer based and patch-based approaches for predicting asset prices and directional price movements. The employed transformer models include Vision Transformer (ViT), Data Efficient Image Transformers (DeiT), and Swin. We use ConvMixer for a patch embedding-based convolutional neural network architecture without a transformer. Our tested transformer-based and patch-based methodologies aim to predict asset prices and directional movements using historical price data by leveraging the inher ent image-like properties within the historical time-series dataset. Before the imple mentation of attention-based architectures, the historical time series price dataset is transformed into two-dimensional images. This transformation is facilitated through the incorporation of various common technical financial indicators, each contribut ing to the data for a fixed number of consecutive days. Consequently, a diverse set of two-dimensional images is constructed, reflecting various dimensions of the dataset. Subsequently, the original images depicting market valleys and peaks are an notated with labels such as Hold, Buy, or Sell. According to the experiments, trained attention-based models consistently outperform the baseline convolutional architec tures, particularly when applied to a subset of frequently traded Exchange-Traded Funds (ETFs). This better performance of attention-based architectures, especially ViT, is evident in terms of both accuracy and other financial evaluation metrics, particularly during extended testing and holding periods. These findings underscore the potential of transformer-based approaches to enhance predictive capabilities in asset price and directional forecasting., Derin ¨o˘grenme metodolojileri tarafından y¨onlendirilen algoritmik ticaret alanı son za manlarda b¨uy¨uk ilgi g¨orm¨u¸st¨ur. Bu alanda, d¨on¨u¸st¨ur¨uc¨uler, evri¸simli sinir a˘gları ve yama g¨omme tabanlı teknikler bilgisayarla g¨orme toplulu˘gu i¸cinde pop¨uler se¸cenekler olarak ortaya ¸cıkmı¸stır. Burada, en yeni bilgisayarla g¨orme metodolojilerinden ve zaman serisi veri k¨umeleri i¸cin g¨or¨unt¨u benzeri d¨on¨u¸s¨um yetene˘gini g¨osteren mev cut ¸calı¸smalardan esinlenerek, varlık fiyatlarını ve y¨onl¨u fiyat hareketlerini tahmin etmek i¸cin daha geli¸smi¸s d¨on¨u¸st¨ur¨uc¨u tabanlı ve yama tabanlı yakla¸sımlar uyguluy oruz. Kullanılan d¨on¨u¸st¨ur¨uc¨u modelleri arasında Vision Transformer (ViT), Data Efficient Image Transformers (DeiT) ve Swin bulunmaktadır. D¨on¨u¸st¨ur¨uc¨us¨uz yama g¨omme tabanlı bir evri¸simsel sinir a˘gı mimarisi i¸cin ConvMixer kullanıyoruz. Test edilen d¨on¨u¸st¨ur¨uc¨u tabanlı ve yama tabanlı metodolojilerimiz, ge¸cmi¸s zaman serisi veri k¨umesindeki do˘gal g¨or¨unt¨u benzeri ¨ozelliklerden yararlanarak ge¸cmi¸s fiyat veri lerini kullanarak varlık fiyatlarını ve y¨on hareketlerini tahmin etmeyi ama¸clamaktadır. Dikkat tabanlı mimarilerin uygulanmasından ¨once, tarihsel zaman serisi fiyat veri seti iki boyutlu g¨or¨unt¨ulere d¨on¨u¸st¨ur¨ul¨ur. Bu d¨on¨u¸s¨um, her biri sabit sayıda ardı¸sık g¨un i¸cin verilere katkıda bulunan ¸ce¸sitli ortak teknik finansal g¨ostergelerin dahil edilme siyle kolayla¸stırılmı¸stır. Sonu¸c olarak, veri setinin ¸ce¸sitli boyutlarını yansıtan farklı bir iki boyutlu g¨or¨unt¨u seti olu¸sturulmu¸stur. Daha sonra, piyasa vadilerini ve zirvelerini g¨osteren orijinal g¨or¨unt¨ulere Tut, Al veya Sat gibi etiketler eklenmi¸stir. Deneylere g¨ore, e˘gitilmi¸s dikkat tabanlı modeller, ¨ozellikle sık i¸slem g¨oren Borsa Yatırım Fon larının (ETF’ler) bir alt k¨umesine uygulandı˘gında, temel konvol¨usyonel mimariler den s¨urekli olarak daha iyi performans g¨ostermektedir. Dikkat tabanlı mimarilerin, ¨ozellikle de ViT’nin bu daha iyi performansı, ¨ozellikle uzun test ve bekletme s¨ureleri boyunca hem do˘gruluk hem de di˘ger finansal de˘gerlendirme ¨ol¸c¨utleri a¸cısından be lirgindir. Bu bulgular, d¨on¨u¸st¨ur¨uc¨u tabanlı yakla¸sımların varlık fiyatı ve y¨on tah mininde ¨ong¨or¨u yeteneklerini geli¸stirme potansiyelinin altını ¸cizmektedir.

View in source Özyeğin University Özyeğin University - Historical works, archives, and periodicals search engine

Özyeğin University

Deep transformer-based asset price and direction prediction

Author Gezici, Abdul Haluk Batur

Publication Date 2024

Subject Deep learning (Machine learning), Cryptocurrencies, Bitcoin, Finance, Mathematical models, Artificial intelligence, Financial applications

Type Document

Language English

Digital Yes

Manuscript No

Library Özyeğin University

Record ID 9637baaa-9e29-46be-adb2-c4b9735262b7

Library Location Department of Artificial Intelligence

Date 2024

Sample Text The field of algorithmic trading, driven by deep learning methodologies, has garnered substantial attention in recent times. Within this domain, transformers, convolutional neural networks, and patch embedding-based techniques have emerged as popular choices within the computer vision community. Here, inspired by the latest cutting edge computer vision methodologies and the existing work showing the capability of image-like conversion for time-series datasets, we apply more advanced transformer based and patch-based approaches for predicting asset prices and directional price movements. The employed transformer models include Vision Transformer (ViT), Data Efficient Image Transformers (DeiT), and Swin. We use ConvMixer for a patch embedding-based convolutional neural network architecture without a transformer. Our tested transformer-based and patch-based methodologies aim to predict asset prices and directional movements using historical price data by leveraging the inher ent image-like properties within the historical time-series dataset. Before the imple mentation of attention-based architectures, the historical time series price dataset is transformed into two-dimensional images. This transformation is facilitated through the incorporation of various common technical financial indicators, each contribut ing to the data for a fixed number of consecutive days. Consequently, a diverse set of two-dimensional images is constructed, reflecting various dimensions of the dataset. Subsequently, the original images depicting market valleys and peaks are an notated with labels such as Hold, Buy, or Sell. According to the experiments, trained attention-based models consistently outperform the baseline convolutional architec tures, particularly when applied to a subset of frequently traded Exchange-Traded Funds (ETFs). This better performance of attention-based architectures, especially ViT, is evident in terms of both accuracy and other financial evaluation metrics, particularly during extended testing and holding periods. These findings underscore the potential of transformer-based approaches to enhance predictive capabilities in asset price and directional forecasting., Derin ¨o˘grenme metodolojileri tarafından y¨onlendirilen algoritmik ticaret alanı son za manlarda b¨uy¨uk ilgi g¨orm¨u¸st¨ur. Bu alanda, d¨on¨u¸st¨ur¨uc¨uler, evri¸simli sinir a˘gları ve yama g¨omme tabanlı teknikler bilgisayarla g¨orme toplulu˘gu i¸cinde pop¨uler se¸cenekler olarak ortaya ¸cıkmı¸stır. Burada, en yeni bilgisayarla g¨orme metodolojilerinden ve zaman serisi veri k¨umeleri i¸cin g¨or¨unt¨u benzeri d¨on¨u¸s¨um yetene˘gini g¨osteren mev cut ¸calı¸smalardan esinlenerek, varlık fiyatlarını ve y¨onl¨u fiyat hareketlerini tahmin etmek i¸cin daha geli¸smi¸s d¨on¨u¸st¨ur¨uc¨u tabanlı ve yama tabanlı yakla¸sımlar uyguluy oruz. Kullanılan d¨on¨u¸st¨ur¨uc¨u modelleri arasında Vision Transformer (ViT), Data Efficient Image Transformers (DeiT) ve Swin bulunmaktadır. D¨on¨u¸st¨ur¨uc¨us¨uz yama g¨omme tabanlı bir evri¸simsel sinir a˘gı mimarisi i¸cin ConvMixer kullanıyoruz. Test edilen d¨on¨u¸st¨ur¨uc¨u tabanlı ve yama tabanlı metodolojilerimiz, ge¸cmi¸s zaman serisi veri k¨umesindeki do˘gal g¨or¨unt¨u benzeri ¨ozelliklerden yararlanarak ge¸cmi¸s fiyat veri lerini kullanarak varlık fiyatlarını ve y¨on hareketlerini tahmin etmeyi ama¸clamaktadır. Dikkat tabanlı mimarilerin uygulanmasından ¨once, tarihsel zaman serisi fiyat veri seti iki boyutlu g¨or¨unt¨ulere d¨on¨u¸st¨ur¨ul¨ur. Bu d¨on¨u¸s¨um, her biri sabit sayıda ardı¸sık g¨un i¸cin verilere katkıda bulunan ¸ce¸sitli ortak teknik finansal g¨ostergelerin dahil edilme siyle kolayla¸stırılmı¸stır. Sonu¸c olarak, veri setinin ¸ce¸sitli boyutlarını yansıtan farklı bir iki boyutlu g¨or¨unt¨u seti olu¸sturulmu¸stur. Daha sonra, piyasa vadilerini ve zirvelerini g¨osteren orijinal g¨or¨unt¨ulere Tut, Al veya Sat gibi etiketler eklenmi¸stir. Deneylere g¨ore, e˘gitilmi¸s dikkat tabanlı modeller, ¨ozellikle sık i¸slem g¨oren Borsa Yatırım Fon larının (ETF’ler) bir alt k¨umesine uygulandı˘gında, temel konvol¨usyonel mimariler den s¨urekli olarak daha iyi performans g¨ostermektedir. Dikkat tabanlı mimarilerin, ¨ozellikle de ViT’nin bu daha iyi performansı, ¨ozellikle uzun test ve bekletme s¨ureleri boyunca hem do˘gruluk hem de di˘ger finansal de˘gerlendirme ¨ol¸c¨utleri a¸cısından be lirgindir. Bu bulgular, d¨on¨u¸st¨ur¨uc¨u tabanlı yakla¸sımların varlık fiyatı ve y¨on tah mininde ¨ong¨or¨u yeteneklerini geli¸stirme potansiyelinin altını ¸cizmektedir.