Studies on computational intelligence based strategies for financial portfolio optimization

Abstract

A financial portfolio is a basket of tradable assets such as stocks, bonds, commodities etc., that is held by an investor. Computational Intelligence (CI) is a consortium of nature inspired computational methodologies and strategies that have proved to be efficient in solving problems on which traditional methods of solution had been rendered ineffective or infeasible. This thesis broadly deals with studies on CI based strategies for the solution of complex constrained portfolio optimization problems. The constraints that have been considered for inclusion in the mathematical model are one or more combinations of basic, bounding, cardinality, class, short sales and transaction costs constraints. The specific CI based methodologies considered in the work are neural networks, wavelet networks and evolutionary strategies. The studies undertaken in the thesis have been viewed under four segments. In the first part of the work, studies on obtaining a better noise filter for the estimation of the empirical covariance matrix, which is one of the key inputs to the constrained portfolio optimization problem have been undertaken. The second part of the work pertains to the solution of a complex constrained portfolio optimization problem. In the third part, a new hybrid strategy named Pareto-archived Evolutionary Wavelet Network (PEWN) is proposed to solve the constrained multi objective portfolio optimization problem. In the last segment, the need for the inclusion of transaction costs in the multi-period portfolio rebalancing problem has been studied. All the aforementioned hybrid strategies have been implemented using MATLAB and demonstrated on the Bombay Stock Exchange (BSE200 index: July 2001 to July 2006) and Tokyo Stock Exchange (Nikkei225 index: March 2002 to March 2007) data sets. newline newline newline