摘要： 利用粒子群及其改进算法，快速精确地估计ARCH 模型的参数，动态地度量描述证券市场收益序列的条件异方差；并利用算法建立美国证券市场道琼斯指数收益的ARCH 模型，进行了走势预测。得到了大致跟随指数的实际走势的预测值，说明ARCH 模型确实能够描述证券市场的“异方差”现象。

关键词: ARCH 模型；PSO 算法；异方差；惯性权重法；压缩因子法

Abstract: This paper estimates the parameters in ARCH model accurately with particle swarm optimization and propose its improved approaches,dynamic depict “heteroskedasticity” of stock return. And the ARCH models for Dow-Jones average stock return are established with algorithm and forecast of the return is given. Actual foreacast value following the index is obtained, which proves that the ARCH model can truly describe the“heteroskedasticity” in the stock market.

Key words: ARCH model; PSO algorithm; Heteroskedasticity; Inertia weight algorithm; Constriction factor algorithm