Mutual fund portfolio optimization using clustering and particle swarm optimization: evidence from Indian markets
TL;DR
A heuristic framework that combines correlation-based hierarchical clustering with Particle Swarm Optimization to construct a diversified fund-of-funds portfolio comprising equity mutual funds demonstrates comparatively strong performance relative to Mean–Variance, Minimum-Variance, Equal-Weight and benchmark portfolios.
Abstract
The large and correlated universe of mutual fund schemes poses challenges in identifying representative funds and constructing efficient portfolios. This study proposes a heuristic framework that combines correlation-based hierarchical clustering with Particle Swarm Optimization (PSO) to construct a diversified fund-of-funds portfolio comprising equity mutual funds. Clustering reduces dimensionality and enhance diversification by selecting representative schemes from a large investment universe. Portfolio weights are then obtained using PSO under long-only, full-investment and bounded-weight constraints, incorporating transaction costs and periodic rebalancing. Using a seven-year estimation period and a three-year out-of-sample evaluation window, the optimized portfolio demonstrates comparatively strong performance relative to Mean–Variance, Minimum-Variance, Equal-Weight and benchmark portfolios, with higher Sharpe, Sortino and Treynor ratios and a positive Jensen’s alpha, without a commensurate increase in volatility or drawdown. The evidence from Indian equity mutual funds indicates that the proposed framework provides a competitive and implementable approach to mutual fund portfolio construction.