基于双重遗传算法的动态模糊系统设计与优化

Design and optimization of dynamic fuzzy system based on double-layer genetic algorithms

设计了一种动态模糊系统模型,该模型能够动态地进行规则自学习,从而减少了规则学习的计算量。提出优化该模型的双重遗传算法:外层采用整数编码,用来训练系统的结构,内层采用实数编码,用来训练系统的参数;内层GA得到的最佳染色体适应值用来评价外层GA相应染色体。该模型结构简单,便于实现,并可离线优化,在线预测。通过应用于实际股市行情的预测和效率分析,不论从预测的结果还是从运行效率看,都收到了满意的效果。

A dynamic fuzzy system forecast model is proposed. The mode can dynamically learn rules by if self and reduce the computational load of rules learning. Then a double-layer genetic algorithm which optimizes the model is proposed. In the algorithm, the external-layer adopts integer codes to train the structure of system and the internal-layer adopts real codes to train the parameters of system. The fitness value of the optimal chromosome from internal-lays GA is used to evaluate corresponding chromosomes of external-layer GA. This model has a simple structure and is easy to implement, and it can be optimized off-line and forecast on-line. Through its application to actual stock market forecast and efficiency analysis, the model is satisfactory both in forecast results and in work efficiency.