Index optimization replication algorithm by using the soft subspace clustering method

This paper proposes a new index optimization replication algorithm framework. First of all, by using independent component analysis technology to build time series feature subspace, we can convert the observation data, which is high dimensional dynamic time series, into static data. Then, use soft subspace clustering method to achieve fuzzy feature weighted clustering. Finally, minimize tracking error and determine the weights of component stocks in the index tracking portfolio. This way, we complete index optimization of replication. The replication method proposed in this paper proves to be effective by positive analysis of China´s CSI 300 index optimization replication.