Bacterial Foraging Optimization with Neighborhood Learning for Dynamic Portfolio Selection

摘要

This paper proposes a new variant of bacterial foraging optimization, called Bacterial Foraging Optimization with Neighborhood Learning (BFONL). In the proposed BFO-NL, information sharing among each individual can be realized by using a von Neumann-style neighborhood topology. To demonstrate the efficiency of BFO-NL in dealing with real world problem, this paper improves the original mean-variance portfolio model into Two-Period dynamic PO model considering risky assets for trading, then uses BFO-NL to automatically find the optimal portfolios in the advanced model. With a five stock portfolio example, BFO-NL is proved to outperform original BFO in selecting optimal portfolios.