New Concepts in Financial Forecasting : Network-Based Information, Topological Data Analysis and their Combination

This paper introduces novel financial predictors that are derived from the interaction profile of financial markets. These predictors utilize network-based and topological information. Since these predictors are derived from the inner dynamics (microstructure) of financial markets, they can be best described as microstructural predictors. After equipping the reader with the methodological background of the novel predictors, we perform an extensive in-sample and out-of-sample performance analyses. The in-sample studies demonstrate that microstructural predictors and their combinations are informative with regard to future asset returns. In the out-of-sample studies, we combine microstructual predictors with state of the art machine learning and statistical factor extraction methods. The resulting active forecasting models dominate the benchmark mean model in terms of profitability, but not it terms of statistical precision. Since an investor is usually more concerned with profitability of active investment strategies, the out-of-sample results confirm the value-added of the novel predictors