基于遗传规划的相关关系预测方法研究

基于遗传规划的搜索寻优技术,给出了一种经济变量与其解释变量间相关关系的预测方法。该方法避开了统计中假设、变换的影响,可以获得显性预测函数表达式,能够处理回归分析法难以解决的非线性问题,比较适合经济变量间复杂关系的分析研究。计算过程中结合蒙特卡洛优化法改进了遗传规划的初始群体,从而提高了进化的全局收敛速度。同时,计算实例表明了预测结果的合理性。

Study on prediction method for related relation based on genetic programming

According to searching and optimization techniques of Genetic Programming, this paper presents a prediction method for related relation between economic indices and influence factors that needn't statistical assumption and transformation, can get prediction function, can deal with no-linear problem which is hard to solve with regression method, as well as can be used to analyze complex relation between economic variables. This paper also improves convergence efficiency of Genetic Programming by means of applying Monte Carlo optimization method to generate initial population. A practical example given in this paper shows that the prediction result is feasible.