基于SVM预测模型的汽车电子市场价值估计

摘要

To evaluate the potential value of automobile electronics market, a prediction model of monthly Chinese automobile production is introduced based on the improved Support Vector Machine (SVM) which is incorporated with the optimized kernel parameters. The SVM model uses RBF kernel function and e-SVR regression method, where the parameters selection problem boils down to an optimization problem of generalization capability minimization and dual problem maximization. According to the monthly Chinese automobile production data in 2005?009, the monthly Chinese automobile production in the former three months of 2010 is predicted, and the potential value of Chinese light vehicle electronics market is evaluated. The result shows that the model can improve the short term prediction performance, which provides a valuable reference for the decision-making of motor companies.%为估计汽车电子市场的潜在价值,引入一种基于改进优化核函数参数支持向量机( Support Vector Machine,SVM)的中国汽车月产量预测模型.SVM采用RBF核函数和ε-SVR回归方法；参数选择归结为使推广能力的估计值最小、对偶问题最大化的最优化问题.根据2005-2009年中国汽车月产量数据,预测2010年前3个月的中国汽车月产量,并估计中国轻型汽车电子市场的潜在价值.结果表明:该模型能够提高短期预测性能,可为汽车公司的市场决策提供有价值的参考.