Research of Pension Fund Market Risk Model Based on Data Mining

Brought forward a novel algorithm to measure VAR based on data mining, and considered of decay and magnify attribute of financial time series to optimize risk market model. First, VAR estimation model was established with the thought of quantile plot, and different time segment´s VAR was calculated under given confidence level. Secondly, VAR´s Failure Frequency was got statistically according to portfolio´s real profit or loss value, which is used to construct the discriminant of the best decay and magnify factor. Finally VAR was gained. This novel algorithm was adopted by Chinese Social Security Fund invest management and control system. The experiment results show that the VAR´s Failure Frequency is between 2.65%-5.56% under given confidence level 95%, is close to 5% and the algorithm is accurate and reliable.