: Traditional integer differentiation (like computing returns) removes "memory" from data. Fractional differentiation aims to achieve stationarity while preserving as much memory as possible.
The field of (FinML) has moved beyond simple predictive models, largely influenced by Marcos López de Prado's seminal work, Advances in Financial Machine Learning . This discipline addresses the unique challenges of financial data, such as low signal-to-noise ratios and non-IID (Independent and Identically Distributed) properties. Core Methodologies in Modern FinML Advances in Financial Machine Learning
: Standard cross-validation fails in finance due to data leakage. These techniques remove overlapping or correlated observations to ensure the model isn't "cheating" by looking at the future. This discipline addresses the unique challenges of financial
Professional fund management requires solving systemic hurdles that often cause retail ML projects to fail: Tommylee1013/Advances-in-Financial-Machine-Learning or time limit.
: Moving away from standard time-based bars to Tick , Volume , or Dollar bars helps synchronized data with market activity levels.
: Using a second ML model to decide whether to act on the primary model's prediction, effectively acting as a "size" or "filter" layer to reduce false positives. Feature Engineering :
: A sophisticated labeling technique that classifies observations based on whether they hit a profit take, stop loss, or time limit.