Stock index tracking by Pareto efficient genetic algorithm

This paper proposes a heuristic searching approach on construction of a tracking portfolio, which is able to get the average market return and can even outperform some hedge funds that are managed actively. The tracking portfolio is expected to replicate the performance of a benchmark index return with a part of its component stocks while reducing the cost of transaction by limiting the number of rebalancing and unnecessary investment on less influential component stocks. The mathematical model being proposed is based on a hybrid genetic algorithm with a self-adaptive evolving mechanism. In order to enhance the model efficiency, we optimize the original genetic algorithm by applying Pareto efficiency as utility measure and goal programming for the inevitable conflicts of multiple objectives/interests. The proposed approach provides a comprehensive solution to index tracking problem by considering as many practical issues as possible. The constructed portfolio has a satisfactory performance on experiments based on CSI300, FTSE100 and HSI data. The proposed formulation of index tracking is therefore believed to be a good alternative to many current techniques.

Figure optionsDownload as PowerPoint slideHighlights
• A tracking portfolio is built by a hybrid genetic algorithm.
• The original heuristic learning is optimized by the principal of Pareto efficiency and goal programming.
• The market constraints are reflected in model specifications.
• CSI300, FTSE100, HSI data are investigated and modeled using the proposed algorithm.