PulseAugur / Brief
EN
LIVE 12:46:35

Brief

last 24h
[2/2] 222 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

  1. TOOL · arXiv cs.LG English(EN) ·

    Machine Learning-Based Bitcoin Trading Under Transaction Costs: Evidence From Walk-Forward Forecasting

    A new research paper explores the use of machine learning models, including XGBoost, LSTM, and iTransformer, for predicting Bitcoin returns. The study found that while these models can generate positive gross trading performance, profitability is lost once transaction costs are considered. However, a cost-aware execution filter that only trades when forecast magnitude exceeds a threshold restored profitability in some configurations, with one XGBoost strategy yielding over 65% annualized returns. AI

    IMPACT Demonstrates potential for ML models in financial markets, but highlights transaction costs as a key challenge for profitability.

  2. RESEARCH · arXiv stat.ML English(EN) ·

    Early Detection of Latent Microstructure Regimes in Limit Order Books

    Researchers have developed a novel method for detecting latent stress regimes in limit order books, which can precede observable market instability. Their approach utilizes a three-regime causal data-generating process to identify a prediction window before stress becomes apparent. A trigger-based detector combining MAX aggregation, a rising-edge condition, and adaptive thresholding demonstrated a mean lead-time of over 18 timesteps in simulations, outperforming traditional baselines. AI

    Early Detection of Latent Microstructure Regimes in Limit Order Books

    IMPACT Introduces a new predictive model for financial market stress, potentially improving algorithmic trading strategies.