【摘要】 在分析现有城市道路交通量短时预测方法缺陷的基础上,针对目前广泛采用的基于经验风险最小化的BP网络易于陷入局部最优解等缺点,结合遗传算法容易寻找全局最优解与支持向量机回归法具有结构风险最小化的特点,提出了将两种算法相结合的GSVMR预测模型,该模型同时具有结构风险最小和容易寻找最优解的双重特性,并对某城市四车道主干道路8∶00~8∶45的交通量进行了预测,结果表明用该模型进行城市道路交通量短时预测所得结果误差较小,依此验证了用GSVMR模型进行城市道路交通量短时预测的有效性。更多还原

【Abstract】 Based on experimental risk minimization and easily plunge into local minimization in forecasting the city road traffic volume,some existing problems of the present short-term traffic volume forecast method such as BP network were analyzed.Then a GSVMR model combining the genetic algorithm and support vector machine regression was established because they have the benefits of finding the overall minimum and structural risk minimization respectively.And a four-lane city road’s traffic volume from 8∶00 to 8∶45 was forecasted by this model as an example.The result shows that the error is very small,thus the validity of GSVMR model was validated.更多还原