PUB - Publikationen an der Universität Bielefeld

An adaptation of the random forest algorithm for Fr & eacute;chet regression is revisited, addressing the challenge of regression with random objects in metric spaces. To overcome the limitations of previous approaches, a new splitting rule is introduced, substituting the computationally expensive Fr & eacute;chet means with a medoid-based approach. The asymptotic equivalence of this method to Fr & eacute;chet mean -based procedures is demonstrated, along with the consistency of the associated regression estimator. This approach provides a sound theoretical framework and a more efficient computational solution to Fr & eacute;chet regression, broadening its application to nonstandard data types and complex use cases.