The Research On Variable Selection With Particle Swarm Optimization Algorithm Based On EV Model

Due to the stock index and the stock index futures developed greatly in the stock market in China, there are more and more arbitrage opportunities with stock index and stock index futures which are hot spots and key focuses in the field of quantitative re-search in China’s securities market. Recently many arbitrage methods have made great success which hedged stock index futures and their underlying stocks. The stock index reflects the overall features of the stock market which composed by a certain number of shares, so it is a high dimensional data, if we use traditional regression models to analyze it, we will suffer from many problems. But if we use variable selection models to select part of constituent stocks of the stock index to make arbitrage with the stock index futures, we will not only avoid large cost fees and the curse of dimensionality, but also make convenience to trade in the real market. There are some measurement errors in the observation variables, and if we ignore them, the estimated parameters will be biased and incongruence.In order to solve the practical problems in the field of financial quantization more effectively, this paper introduces a kind of variable selection method combined particle swarm optimization algorithm based on an EV model. Considering the good perfor-mance of the Lasso and adaptive Lasso variable selection models, this paper combines the penalty functions of the Lasso and adaptive Lasso models to estimate parameters. Through numerical simulation to prove our model having good performance. Finanlly we use the IH 50, IF 300 and their constituent stocks to make empirical analysis. This model not only solves the problems of traditional variable selection methods that are not applying to EV models, but also has higher accuracy and faster convergence speed.