A partitioned portfolio insurance strategy by a relational genetic algorithm

Abstract

This paper proposes a new portfolio insurance (PI) strategy named partitioned portfolio insurance (PPI) strategy and a new relation-based genetic algorithm named relational genetic algorithm (RGA) to optimize the proposed PPI strategy. Our PPI strategy extends the traditional PI strategy to become a more aggressive one. In our PPI strategy, we attempt to correctly partition the portfolio into several similar sub-portfolios and then insure the sub-portfolios individually. It not only avoids the downside risk, but also further explores the upside profit successfully. In addition, our RGA which adopts a relational encoding and has a set of problem-independent operators is designed to solve the induced portfolio partitioning problem. The relational encoding eliminates the redundancy of previous GA representations for partitioning problems and improves the performance of genetic search. The problem-independent operators we redesigned manipulate the genes without requiring specific heuristics in the process of evolution. Moreover, our RGA works without requiring a preset number of subsets in advance. Experiments for developing optimized PPI strategies by RGA are performed. Experimental results show that our optimized PPI strategies are significantly better than the traditional PI strategy and our RGA works well for solving the portfolio partitioning problem.