针对基于GPS／GIS的浮动车数据特点，总结其中无效的数据类型，并给出数据有效处理的方法。以支持向量机原理、交通状态预测方法为基础，分析了常用支持向量回归机、核函数及模型参数的性能，以及各核函数及模型参数对支持向量机性能的影响及作用。针对路段平均速度预测中的小样本、非线性、高维回归等特点，将支持向量回归机方法引入基于浮动车数据的路段车辆速度预测，构建了路段平均速度短时预测模型。并以杭州市某路段的实际数据为例，详细阐述了支持向量回归机预测模型的具体建模和求解过程。运用LibSVM2．84软件包，进行预测模型的参数选择、样本训练以及预测求解，并通过预测结果的对比分析，验证了预测模型的可用性和有效性。-Characteristics of the GPS/GIS-based floating car data, summary of which types of invalid data, and gives the effective data processing method. Support vector machine, the traffic state prediction method based on analysis of commonly used support vector regression machines, nuclear function and performance of the model parameters, as well as the kernel function and model parameters on the performance of SVM and effect. Small sample, nonlinear, high dimensional regression for the prediction of average speed characteristics, the support vector regression machine to the introduction of prediction based on floating car data section of the speed of vehicles, build a short-term forecasting model of average speed. In Hangzhou section of the actual data, for example, elaborated on the specific modeling and solving process of the support vector machine for regression prediction model. Use LibSVM2.84 package forecast model parameter selection, sample training, and forecasting solution, and verif