Rule Extraction from Large Numerical Datasets: Construction of ESG Signals
C. Geissler, Advestis and Sustainalytics, December 2016.
Abstract: The investment industry has not waited for the Big Data era to exploit price patterns hidden in historical data. However, the explosion of computational power over the past decade, along with an intense focus on machine learning, has created a completely new framework for data analysis. Today, the proliferation of data sources has become an opportunity for organizations. At the beginning of the 21st century, machine learning as well as environmental, social, and governance (ESG) research was starting to be used more frequently by asset managers worldwide. However, synergies between the two areas had not been effectively explored. To-date, ESG research has primarily been used as part of a qualitative process for risk mitigation, and big data techniques have not been applied to large sets of ESG information. Yet, over the last few years, asset managers have been seeking ways to integrate ESG research into their quantitative models to extract additional alpha and beta sources.