This talk provides an overview of the Learned Sectors project, which presents a new, objective data-driven approach to market sectorization based on company fundamentals. The talk covers the trained hierarchical clustering model used to generate a set of candidate sector universes and the reIndexer validation system used to evaluate and rank the universes. The talk also discusses the final risk-adjusted return optimal sector universe identified through the evaluation process.
Overall, the talk highlights the need for more objective, data-driven approaches to market sectorization and presents a novel method for achieving this. The approach is based on company fundamentals and uses a hierarchical clustering model to generate candidate sector universes. These universes are then evaluated and ranked using a backtest-driven system to identify the optimal sector universe in terms of risk-adjusted return. The talk provides an overview of this process and its results.