Use this URL to cite or link to this record in EThOS:
Title: Investigating portfolio insurance strategies including applications of heuristic optimisation and evolutionary artificial neural networks
Author: Khuman, Anil
ISNI:       0000 0004 2738 0824
Awarding Body: University of Essex
Current Institution: University of Essex
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
This work examines the Constant Proportion Portfolio Insurance (CPPI) investment strategy under the assumption of a cumulative prospect theory (CPT) investor. It is found that such an investor who alters their reference point, dependent on the underlying risky asset, prefers strategies that extend the functionality of the CPP!. These other strategies either modify the floor or multiplier components of the CPPI to protect greater levels of interim wealth, or allow for faster allocation or deallocation of capital. Furthermore, it is established, with regards to transaction costs, that an investor benefits greater from the use of multiplier bounds rebalancing triggers than calendar rebalancing. From the perspective of the seller, managing gap risk is of primary importance. This is the risk that the portfolio value will be below the guarantee at maturity. The impact of this to the seller is a loss of fee income and risk to reputation. An in-depth analysis through simulations under a GARCH price process reveals the gap risk to be modest in most cases and easily covered by realistic fee rates. To facilitate a fair comparison of the performance of the different strategies, their optimal parameter settings must be found. Such problems pose difficulties for traditional optimisation algorithms, motivating the use of heuristic methods. Differential evolution, an evolutionary algorithm, is adopted for this task. Although examining extensions to the CPPI model provides a valuable insight into which components of the model are most beneficial to a CPT investor, the formulations of the extensions themselves may be considered arbitrary. Ideally it should be possible, given the goal of maximising the cumulative prospect value, that an entirely new model be generated that trades optimally for this purpose. It is proposed that an evolutionary artificial neural network (EANN) is a suitable approach. As a initial fundamental milestone to achieve this, a supervised learning problem is formulated with the aim of evolving an ANN to replicate the CPPI strategy and the augmented versions presented. It is demonstrated that an EANN is capable of reproducing the trading output of the discrete-time version of the CPPI.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available