Quantitative Finance > Statistical Finance
[Submitted on 30 Mar 2023]
Title:Online Ensemble of Models for Optimal Predictive Performance with Applications to Sector Rotation Strategy
View PDFAbstract:Asset-specific factors are commonly used to forecast financial returns and quantify asset-specific risk premia. Using various machine learning models, we demonstrate that the information contained in these factors leads to even larger economic gains in terms of forecasts of sector returns and the measurement of sector-specific risk premia. To capitalize on the strong predictive results of individual models for the performance of different sectors, we develop a novel online ensemble algorithm that learns to optimize predictive performance. The algorithm continuously adapts over time to determine the optimal combination of individual models by solely analyzing their most recent prediction performance. This makes it particularly suited for time series problems, rolling window backtesting procedures, and systems of potentially black-box models. We derive the optimal gain function, express the corresponding regret bounds in terms of the out-of-sample R-squared measure, and derive optimal learning rate for the algorithm. Empirically, the new ensemble outperforms both individual machine learning models and their simple averages in providing better measurements of sector risk premia. Moreover, it allows for performance attribution of different factors across various sectors, without conditioning on a specific model. Finally, by utilizing monthly predictions from our ensemble, we develop a sector rotation strategy that significantly outperforms the market. The strategy remains robust against various financial factors, periods of financial distress, and conservative transaction costs. Notably, the strategy's efficacy persists over time, exhibiting consistent improvement throughout an extended backtesting period and yielding substantial profits during the economic turbulence of the COVID-19 pandemic.
Current browse context:
q-fin.ST
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.