Enhanced index tracking with multiple time-scale analysis

This paper considers the process optimal strategies for an enhanced index tracking problem. The investment goals are set to achieve a higher return than the benchmark by setting the portfolio's risk profile identical to the primary index risk factors. Return differences between the index and the tracking portfolio are classified as positive and negative series. Multiple time-scale features of each series are extracted by the method of empirical mode decomposition. Then the positive return deviations are modeled by trend-like low frequency behavior and the negative return deviations are modeled by a trendless high frequency behavior. By adopting an immunity-based multi-objective optimization algorithm, the solutions for the process optimal enhanced index tracking are developed. Five data sets drawn from major world markets are adopted to implement our approach. The computational results show the superiority of our model.