Regensburg 2004 – scientific programme

Parts | Days | Selection | Search | Downloads | Help

AKSOE: Physik sozio-ökonomischer Systeme

AKSOE 1: Finanzm
ärkte und Risikomanagement

AKSOE 1.3: Talk

Monday, March 8, 2004, 11:00–11:30, H8

Tackling Mutual Funds Style Analysis with Medoid Clustering and Differential Evolution — •Thiemo Krink¹, Sandra Paterlini², Tommaso Minerva², and Francesco Pattarin² — ¹EVALife Group, Dept. of Computer Science, Univ. of Aarhus, Denmark — ²Dept. of Political Economics, Univ. of Modena and Reggio Emilia, Italy

Mutual funds style analysis requires objective, representative, consistent and empirically testable classification schemes in order to give reliable information to investors and fund managers who are interested in evaluating and comparing different financial products. Institutional classification schemes, when available, do not always provide consistent and representative peer groups of funds. In this study, we introduce a classification algorithm, which identifies mutual funds styles by analysing the time series of past returns. The proposed classification procedure consists of three steps: (a) dimensionality reduction based on principal component analysis, (b) clustering by a novel medoid evolution approach utilizing differential evolution, and (c) style identification by a constrained regression model. We tested the algorithm regarding Italian mutual funds data and achieved satisfactory results with respect to the agreement with the existing institutional classification and the explanatory power of out of sample variability in the cross-section of returns.