Abstract: In order to improve the predicting accuracy oil price, a novel oil price predicting model is proposed based on unified solving parameters phase space reconstruction and predicting algorithm according to relation between phase space reconstruction and predicting algorithm. The least square support vector machine is selected as the predicting algorithm of oil price, and the parameters of phase space reconstruction and LSSVM are taken as individuals of the genetic algorithm, and the optimal delay time, embedding dimension and LSSVM parameters are obtained through selection, crossover and mutation evolutionary mechanism, the predicting model of oil prices is established and the performance of predicting model is tested by Daqing oil price time series. The results show that the proposed model obtains higher predicting accuracy than the models which phase space reconstruction and LSSVM are optimized independently, and it provides a new research idea for the predicting problem of chaotic time series.

Key words: oil price, least squares support vector machine, phase space reconstruction, unified solving, modeling and predictions

摘要： 为了提高石油价格预测精度，利用相空间重构和预测算法参数之间的相互联系，提出一种基于相空间重构和预测算法参数统一求解的石油价格预测模型（PSR-LSSVM）。选择最小二乘支持向量机作为石油价格预测算法，将相空间重构和LSSVM参数作为遗传算法的个体，通过选择、交叉和变异等进化机制找到最优的时间延迟、嵌入维和LSSVM参数，根据最优参数建立石油价格预测模型，并通过大庆石油价格时间序列对模型性能进行测试。结果表明，相对于独立优化相空间重构和LSSVM参数的石油价格预测模型，PSR-LSSVM获得了更高的石油价格预测精度，为具有混沌性的时间序列预测问题提供了一种新的研究思路。

关键词: 石油价格, 最小二乘支持向量机, 相空间重构, 统一求解, 建模预测