Abstract: An improved dynamic neighborhood particle swarm optimization (IDNPSO) was proposed. A new topological structure which constructs dynamic neighbors was induced， and the parameter settings were dynamically adjusted. To improve the predictive accuracy of BP neural network， an improved prediction method of optimized BP neural network based on IDNPSO was introduced. Making use of the index price of Shanghai composite index and Shenzhen composite index， comparison of the forecast performance between IDNPSOBP and GABP neural networks was taken. The result showed that INDPSOBP neural network outperformed GABP neural network， and had the ability to forecast stock index price.

Key words: neural network, dynamic neighborhood, particle swarm optimization, stock index, forecasting