ebook img

Solving Cardinality Constrained Portfolio Optimisation Problem Using Genetic Algorithms and Ant ... PDF

130 Pages·2015·2.21 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Solving Cardinality Constrained Portfolio Optimisation Problem Using Genetic Algorithms and Ant ...

Solving Cardinality Constrained Portfolio Optimisation Problem Using Genetic Algorithms and Ant Colony Optimisation Yibo Li A thesis submitted for the degree of Doctor of Philosophy School of Information Systems, Computing and Mathematics, Brunel University 2015 i Abstract In this thesis we consider solution approaches for the index tacking problem, in which we aim to reproduces the performance of a market index without purchasing all of the stocks that constitute the index. We solve the problem using three different solution approaches: Mixed Integer Programming (MIP), Genetic Algorithms (GAs), and Ant-colony Optimization (ACO) Algorithm by limiting the number of stocks that can be held. Each index is also assigned with different cardinalities to examine the change to the solution values. All of the solution approaches are tested by considering eight market indices. The smallest data set only consists of 31 stocks whereas the largest data set includes over 2000 stocks. The computational results from the MIP are used as the benchmark to measure the performance of the other solution approaches. The Computational results are presented for different solution approaches and conclusions are given. Finally, we implement post analysis and investigate the best tracking portfolios achieved from the three solution approaches. We summarise the findings of the investigation, and in turn, we further improve some of the algorithms. As the formulations of these problems are mixed-integer linear programs, we use the solver ‘Cplex’ to solve the problems. All of the programming is coded in AMPL. ii Acknowledgement First, I would like to thank my supervisor Dr. Cormac Lucas for his excellent guidance, constant support, and great inspiration through my Ph.D. studies. It was great pleasure to work with him. I would also like to thank my parents and other family members, without their help and support none of these would have been possible. I am very lucky to have all of you stand by my side and I am very grateful for all the things you have done for me. I would like to thank all the staff in the Mathematics Department for their excellent work. Finally, I would also like to thank my Ph.D. colleagues and friends: Cristiano, Xiang, Antonio, Zhenghong, Longhui, Martin, Fei and many others. iii Table of Contents Abstract i Acknowledgement ii Table of Contents iii List of Figures v List of Tables vii 1. Introduction ........................................................................................................................... 1 1.1 Introduction .................................................................................................................. 1 1.2 Thesis Structure ........................................................................................................... 5 2. Literature Review .................................................................................................................. 7 2.1 History of Portfolio Theory ......................................................................................... 7 2.1.1 Mean-Variance Model ...................................................................................... 7 2.1.2 Capital Asset Pricing Model ........................................................................... 11 2.1.3 Factor Model ................................................................................................... 12 2.2 Index Tracking ........................................................................................................... 14 2.2.1 Genetic Algorithm........................................................................................... 14 2.2.2 Other Approaches ........................................................................................... 19 2.3 Summary .................................................................................................................... 22 3. Mixed Integer Programming ............................................................................................... 24 3.1 Introduction ................................................................................................................ 24 3.2 Research Comparisons ............................................................................................... 25 3.3 Formulation ................................................................................................................ 28 3.4 Computational Results ............................................................................................... 30 4. Genetic Algorithms ............................................................................................................. 32 4.1 Introduction ................................................................................................................ 32 4.2 Application for Index Tracking ................................................................................. 33 iv 4.3 The Inverse Triangle Genetic Algorithm ................................................................... 38 4.3.1 Introduction ..................................................................................................... 38 4.3.2 Formulation ..................................................................................................... 39 4.3.3 Simulation ....................................................................................................... 40 4.3.4 Computational Results .................................................................................... 46 4.4 Roulette Genetic Algorithms ..................................................................................... 49 4.4.1 Schema Theorem............................................................................................. 49 4.4.2 Application ...................................................................................................... 53 4.4.3 Investigation of population size ...................................................................... 62 4.4.4 Enhancement ................................................................................................... 65 4.4.5 Computational Results .................................................................................... 66 5. Ant Colony Optimization .................................................................................................... 69 5.1 Introduction ................................................................................................................ 69 5.2 Application ................................................................................................................. 73 5.3 Parameters Investigation ............................................................................................ 82 5.4 Computational Results ............................................................................................... 90 6. Post Analysis ....................................................................................................................... 93 6.1 Tracking Portfolios Investigation .............................................................................. 93 6.1.1 Investigation Method ...................................................................................... 93 6.1.2 Investigation Findings ..................................................................................... 98 6.1.3 Further Application ....................................................................................... 100 6.1.4 Computational Results .................................................................................. 101 6.2 Out-of-sample Performance ..................................................................................... 104 7. Conclusion ........................................................................................................................ 113 7.1 Summary .................................................................................................................. 113 7.2 Contribution to Knowledge ...................................................................................... 115 7.3 Future Ideas and Directions ..................................................................................... 116 Reference .............................................................................................................................. 118 v List of Figures Figure 1.1: Percentage of active managers outperforming their benchmark over rolling five- year periods ······························································································································ 3 Figure 1.2: Percentage of active managers outperforming their benchmark during bull/bear cycles ········································································································································· 3 Figure 2.1: The efficient frontier ······························································································· 9 Figure 2.2: The capital market line ··························································································· 9 Figure 4.1: Crossover ··············································································································34 Figure 4.2: Deformed offspring ·······························································································34 Figure 4.3: Semi-optimization approach ·················································································35 Figure 4.4: Fitness landscape ··································································································36 Figure 4.5: Traditional mutation ·····························································································36 Figure 4.6: Inverse triangle evolution ·····················································································38 Figure 4.7: Evolution with redundant duplicates ····································································44 Figure 4.8: Transformed evolution ·························································································44 Figure 4.9: Portfolio illustration ······························································································54 Figure 4.10: Computing time comparison for f12, f15, and f18 ·············································65 Figure 4.11: Roulette wheel ····································································································66 Figure 5.1: Construction graph ·······························································································70 Figure 5.2: Double bridge experiment ····················································································71 Figure 5.3: Additional experiment ··························································································72 Figure 5.4: Enlargement method ····························································································77 Figure 5.5: Mutation of ACO ···································································································79 Figure 5.6: Extended double bridge experiment ····································································85 Figure 5.7: The results of the extended double bridge experiment ·······································86 Figure 6.1: The return of Hang Seng index and its stocks ·······················································94 Figure 6.2: Half deterministic and half probabilistic searching ············································100 Figure 6.3: Hang Seng, K=10 ·································································································105 Figure 6.4: Hang Seng, K=15 ·································································································105 Figure 6.5: DAX, K=10 ············································································································106 Figure 6.6: DAX, K=15 ············································································································106 vi Figure 6.7: DAX, K=20 ············································································································107 Figure 6.8: FTSE, K=10 ···········································································································107 Figure 6.9: FTSE, K=15 ···········································································································108 Figure 6.10: FTSE, K=20 ·········································································································108 Figure 6.11: S&P, K=10 ··········································································································109 Figure 6.12: S&P, K=15 ··········································································································109 Figure 6.13: S&P, K=20 ··········································································································110 Figure 6.14: Nikkei, K=10 ······································································································110 Figure 6.15: Nikkei, K=15 ······································································································111 Figure 6.16: Nikkei, K=20 ······································································································111 Figure 6.17: Nikkei, K=25 ······································································································112 vii List of Tables Table 3-1: BMC’s MIP data sets ······························································································27 Table 3-2: MIP data sets of our research ················································································27 Table 3-3: Computational results of MIP ················································································31 Table 4-1: IT-GA simulation ·····································································································41 Table 4-2: Initial population size for each index ·····································································46 Table 4-3: Computing results of MIP, IT-GA, and EIT-GA for the first five indices ·················47 Table 4-4: Computational results of EIT-GA and MIP for the larger indices ···························48 Table 4-5 (a): Semi-crossover operator simulation – portfolio processing ····························56 Table 4-6 (a): Semi-optimization operator simulation – portfolio processing ·······················59 Table 4-7: Initial population size under different frequencies ···············································62 Table 4-8: Computational results of R-GA for smaller size indices under different input frequencies ·····························································································································63 Table 4-9: Average Tracking Error ···························································································64 Table 4-10: EIT-GA vs ER-GA ···································································································67 Table 5-1: Simulation ··············································································································76 Table 5-2: Attraction table before amplification ····································································77 Table 5-3: Attraction table after amplification ·······································································77 Table 5-4: Stopping criteria illustration ··················································································81 Table 5-5: Colony size for each index under different frequencies ········································83 Table 5-6: Computational results under different input frequencies ·····································84 Table 5-7: Computational results under different evaporation rates ····································87 Table 5-8: ER-GA vs ACO ·········································································································91 Table 6-1: Close stocks of the Hang Seng index ······································································96 Table 6-2: Optimal portfolio of the Hang Seng index with K=10 and K=15 ····························97 Table 6-3: Tracking portfolios investigation ···········································································98 Table 6-4: Results comparison for GA ···················································································102 Table 6-5: Results comparison for ACO ················································································103 1 Chapter 1 . 1 Introduction 1.1 Introduction Fund managers are often hired to control a large amount of money which is invested in various securities and other assets such as: real estate. The investor can be categorized into two groups: institutions and private investors. Fund managers usually provide investment management services that include financial statement analysis, asset/stock selection, and direction of management plans and ongoing monitoring of investments. The aim for fund managers is to make capital growth and income over the short-term or the long-term. The basic investment strategies adopted by fund managers can be broadly classified into two types: active management and passive management.  Active management For active management, the fund managers have more confidence in their own ability to estimate cash flows, growth rates, and discount rates. Based on these estimates, they value assets and determine whether an asset is fairly valued. In an actively managed portfolio, assets that are undervalued, or have a chance of offering above-normal returns, will have a higher weight than that in the index, whereas other assets will have a zero weight, or even 2 negative weight if short selling is permitted. This style of investing is called active management, and the portfolios are referred to as active portfolios. Most open-end mutual funds and hedge funds practice active management, and most analysts believe that active investment can provide capital growth. Active management often has high fixed cost (payments to the management team) and its high frequency of trading often incurs high transaction costs. If the market performs well these fees will be offset by the returns.  Passive management This is based on the assumption of an efficient market. If the markets are efficient, the price in the market is an unbiased estimate of all future discounted cash flows. In other words, the price aggregates and reflects all information that is publicly available, and investors cannot expect to earn a return that is greater than the required rate of return for that asset. So there is no way to outperform the market. In that case, a simple and convenient approach to investing is to rely on the prices set by the market. Portfolios that are based on the assumption of unbiased market prices are referred to as passive portfolios. Passive portfolios most commonly replicate and track market indices, which are passively constructed on the basis of market prices and market capitalizations. Examples of market indices are the FTSE 100, the Nikkei 225, and the S&P 500. Passive portfolios based on market indices are called index funds and generally have low running costs because no significant effort is expended in valuing securities that are included in an index. By comparison, passive management has lower fixed costs and lower transaction costs, but once the market falls it will inevitably affect the index fund. Historical evidence reveals that active fund managers averagely underperform their corresponding benchmarks. Researchers also found that some of the best active fund managers did perform reasonably well in some periods, but most of them failed to carry their success over a long-term period. The following two figures show the evidence found by researchers from ‘Vanguard’ (Charles Thomas, Peter Westaway and Todd Schlanger, n.d.):

Description:
and Ant-colony Optimization (ACO) Algorithm by limiting the number of stocks that can be held. robust framework to protect against the worst-case realizations of potential estimation errors and other tracking portfolio of the index would incur a large downside risk that could incur large losses f
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.