An Approximation Method of Positive Semi-definite Matrix Based on Weighted F-norm

Positive is the important hypotheses of many financial forecasting models, but the correlation coefficient matrix we get from actual samples are not always positive. Firstly, we introduced how to set correlation coefficient matrix according to the sample, and introduced the theory of norm approximation, based on which we find the proximal correlation coefficient matrix, namely unit diagonal positive semi-definite symmetric matrix. Finally, the validity of method in this paper has been proved in practical experiment.