FinChaPS 1D/2D: Financial Chart Pattern Synthesis
Description
FinChaPS (Financial Chart Pattern Synthesis) is a synthetic benchmark dataset designed for developing, training, and evaluating deep learning models for financial chart pattern recognition and classification. It comprises two complementary datasets: FinChaPS-1D, containing 271,180 one-dimensional synthetic closing-price sequences, and FinChaPS-2D, containing 235,596 corresponding candlestick chart images. Both datasets cover nine financial chart pattern categories, including eight well-known patterns: Single Top (ST), Inverse Single Top (IST), Double Top (DT), Inverse Double Top (IDT), Triple Top (TT), Inverse Triple Top (ITT), Head and Shoulders (HAS), and Inverse Head and Shoulders (IHAS), and a Normal/Other class. The synthetic samples are generated using basic waveform functions and various transformation techniques to introduce diversity and variability. FinChaPS 1D provides 1D numerical time-series representations while FinChaPS 2D includes visual candlestick representations of financial patterns, enabling researchers to investigate and compare different approaches to financial chart pattern recognition using time-series analysis, computer vision, and deep learning techniques.
Files
Institutions
- Imam Khomeini International UniversityQazvin Province, Qazvin