Valuation-Anchored Similarity Learning for Private Markets

A CatBoost-based ensemble framework identifies peer companies in private markets by training on observed valuations rather than static features. This approach uses importance-weighted leaf-node co-occurrences to capture non-linear drivers of company similarity despite sparse data.