Fractional differencing is currently widely explored in finance.
Integer differencing transforms a financial time series by subtracting consecutive observations, and is commonly used to remove trends and make nonstationary data stationary. However, integer differencing often removes too much information and can destroy long-term dependence patterns that are important in financial data. Fractional differencing generalizes this idea by allowing the differencing order to take non-integer values, applying a weighted combination of many past observations rather than a simple first difference. This approach can achieve stationarity while preserving long-memory structure and predictive information embedded in historical data. As a result, fractional differencing often provides a better balance between noise reduction and information retention for financial forecasting and machine learning applications.
Presenter: Samson Lam is an MS candidate in Quantitative Finance at Fordham
University's Gabelli School of Business. He holds undergraduate
degrees in Quantitative Finance and Risk Management Science, with
minors in Mathematics and Statistics, from the Chinese University of
Hong Kong. His research focuses on the discovery and design of novel
statistical methodologies for financial markets, with a primary focus
on financial econometrics.
