Multi-period model of portfolio investment and adjustment based on hybrid genetic algorithm

Abstract

This paper proposes a multi-period portfolio investment model with class constraints, transaction cost, and indivisible securities. When an investor joins the securities market for the first time, he should decide on portfolio investment based on the practical conditions of securities market. In addition, investors should adjust the portfolio according to market changes, changing or not changing the category of risky securities. Markowitz mean-variance approach is applied to the multi-period portfolio selection problems. Because the sub-models are optimal mixed integer program, whose objective function is not unimodal and feasible set is with a particular structure, traditional optimization method usually fails to find a globally optimal solution. So this paper employs the hybrid genetic algorithm to solve the problem. Investment policies that accord with finance market and are easy to operate for investors are put forward with an illustration of application.

Keywords

portfolio / transaction cost / class constraint / hybrid genetic algorithm

References

[1] Markowitz H. M. Portfolio selection[J]. J of Finance, 1952, 7(1): 77-91.

[2] Mossin J. Optimal multiperiod portfolio policies[J]. Journal of Business, 1968, 4l(2): 215-229.

[3] Fama E. E. Multiperiod consumption-investment decisions[J]. American Economic Review, 1970, 60(1): 163-l74.

[4] Merton R. C. Continuous-Time Finance[M]. 1990, Cambridge, MA: Basil Blackwell.

[5] Grauer R. R., Hakansson N. H. On the use of mean-variance and quadratic approximations in implementing dynamic investment strategies: A comparison of returns and investment policies[J]. Management Science, 1993, 39(7): 856-871.

[6] Stein E. F. The performance of stochastic dynamic and fixed mix portfolio model[J]. European Journal of Operational Research, 2003, 140(1): 37-49.

[7] Bjarne H., Michael T. Optimal dynamic portfolio selection for a corporation with controllable risk and dividend distribution policy[J]. Quantitative Finance, 2004, 4(3): 315-327.

[8] Cerny A. Optimal continuous-time hedging with Leptokurtic returns[J]. Mathematical Finance, 2007, 17(2): 175-203.

[9] Robert A., Rogers L. C. G. Valuations and dynamic convex risk measures[J]. Mathematical Finance, 2008, 18(1): 1 22

[10] Li Z., Wang Shouyang. EaR risk measurement and dynamic investment decision[J]. The Journal of Quantitative and Technical Economics, 2003, 16(1): 45-53.

[11] Tang Wansheng. An intelligent solving method for multiperiod investment decision[J]. Journal of Systems Engineering, 2003, 21(2): 120-124.

[12] Li D., Ng W. L. Optimal dynamic portfolio selection: Multiperiod mean-variance formulation[J]. Mathematical Finance, 2000, 10(3): 387-406.

[13] Liu Y., Kang L., Chen Yuping. Non-numerical Parallel Algorithm — Genetic Algorithm[M]. 1998, Beijing: Science Press.

[14] Kang L., Xie Y., You S., et al. Non-numerical Parallel Algorithm — Simulated Annealing[M]. 1998, Beijing: Science Press.

[15] Oestermark R. Solving a non-convex trim loss problem with a genetic hybrid algorithm[J]. Computer and Operation Research, 1999, 26(6): 623-635.

[16] Rong X., An Zhiyu. Hybrid genetic algorithms for the nonlinear programming[J]. Systems Engineering and Electronics, 2003, 25(5): 621-624.