notesum.ai

Published at December 5

Dynamic Graph Representation with Contrastive Learning for Financial Market Prediction: Integrating Temporal Evolution and Static Relations

cs.LG

cs.NE

q-fin.CP

Released Date: December 5, 2024

Authors: Yunhua Pei¹, Jin Zheng¹, John Cartlidge¹

Aff.: ¹University of Bristol

Arxiv: http://arxiv.org/pdf/2412.04034v1

Refer to caption

Methods	NASDAQ dataset			NYSE dataset
Methods	Accuracy(%)	F1 score	MCC( $\times 10^{-3}$ )	Accuracy(%)	F1 score	MCC( $\times 10^{-3}$ )
Graph WaveNet (Wu et al.,, 2019)	50.94±0.15	60.51±1.10	5.50±1.00	50.86±0.13	57.28±1.86	11.12±2.21
MTGODE (Jin et al.,, 2022)	51.41±0.15	62.41±4.24	5.92±2.61	52.35±0.17	62.73±1.72	20.62±5.76
STGCL (Liu et al., 2022b, )	50.56±0.08	57.81±1.19	9.71±1.15	51.21±0.05	67.47±0.07	22.33±1.53
EvolveGCN (Pareja et al.,, 2020)	50.19±0.02	56.94±0.21	4.41±0.76	51.14±0.15	66.85±0.37	14.41±0.45
DGRCL (Ours)	53.06±0.18	66.53±2.74	9.32±1.53	54.07±0.20	67.53±0.19	27.51±5.52