Researchers have developed DDSL, an iterative self-labeling framework designed to construct labels for financial market states. This framework couples a price- and grammar-guided Viterbi labeler with a Transformer classifier, using model feedback to refine labels. Experiments on major stock indices like SSE50, CSI300, CSI500, and NASDAQ100 demonstrated that DDSL significantly improves joint-state macro F1 scores compared to traditional labeling methods. The study also indicated that the learnability of these labels can be controlled by iteration and model feedback, with implications for price-feature separation and conditional reversal signals. AI
IMPACT Introduces a novel iterative self-labeling framework for financial markets, potentially improving predictive accuracy and control over label learnability.
RANK_REASON The cluster describes a new research paper detailing a novel framework for financial market state labeling. [lever_c_demoted from research: ic=1 ai=0.4]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →