基于神经网络和SVM的短时交通流组合预测研究
发布时间：2018-10-17 10:54
【摘要】：短期交通流预测是智能交通的热点研究领域,短期交通流预测是交通控制和车载导航的重要问题之一,也是智能交通控制和交通诱导的关键技术之一。短期交通流具有非线性、时变性、不确定性、不稳定性等特点,以各种预测模型实现短时交通流预测,可以缓解城市交通拥堵,避免社会资源的浪费。因此,研究短时交通流的预测具有重要的现实意义和应用价值。 对于非线性时间序列的交通流预测,采用RBF神经网络具有较好的效果,其预测准确度也较高,但RBF神经网络具有泛化较差的缺点。 为了克服RBF神经网络的缺点,采用支持向量机对交通流进行预测,但是支持向量机仅对200个以下的小样本数据可以获得较好的预测结果,而交通流的数据样本一般在上千个左右,故用支持向量机对交通流进行预测的误差仍然比较大。 为了克服RBF神经网络和支持向量机各自的缺点,本文把这两个模型组合使用,根据不同的组合算法达到不同的组合预测效果。利用权值计算公式对两种预测模型进行权值计算,通过以往预测结果的误差得出两种预测结果的权值,从而得到较高的预测结果。由于这种权值计算公式是基于经验计算的统计结果,其预测误差是随机的、非线性的,于是采用支持向量机来对上述两个模型的预测结果进行二次预测,让支持向量机来计算两种单一预测模型的权值,从而得到了更精确的预测结果。 实验结果表明,采用组合模型的预测比单一模型的预测效果更好。相对地,使用支持向量机计算权值的组合预测模型要比使用权值计算公式的组合预测模型得到的预测结果要精确一些,预测效果更好一些。
[Abstract]:Short-term traffic flow prediction is a hot research field in intelligent transportation. Short-term traffic flow prediction is one of the important problems in traffic control and vehicle navigation, and is also one of the key technologies of intelligent traffic control and traffic guidance. Short-term traffic flow has the characteristics of nonlinearity, time-varying, uncertainty, instability and so on. Using various forecasting models to forecast short-term traffic flow can alleviate urban traffic congestion and avoid the waste of social resources. Therefore, the study of short-time traffic flow prediction has important practical significance and application value. For the traffic flow prediction of nonlinear time series, the RBF neural network has a good effect and its prediction accuracy is higher, but the RBF neural network has the disadvantage of poor generalization. In order to overcome the shortcoming of RBF neural network, support vector machine (SVM) is used to predict traffic flow. The data samples of traffic flow are usually thousands or so, so the error of forecasting traffic flow with support vector machine is still large. In order to overcome the shortcomings of RBF neural network and support vector machine, this paper combines the two models to achieve different combination prediction results according to different combination algorithms. Weight calculation formula is used to calculate the weights of the two prediction models, and the weight values of the two prediction results are obtained by the error of the previous prediction results, and the higher prediction results are obtained. Because this formula is based on the statistical results of empirical calculation and the prediction error is random and nonlinear, support vector machine (SVM) is used to predict the prediction results of the above two models. Support vector machine (SVM) is used to calculate the weights of two single prediction models, and a more accurate prediction result is obtained. The experimental results show that the combined model is more effective than the single model. Comparatively, the combined prediction model using support vector machine to calculate the weight value is more accurate than the combination prediction model with the formula of the right to use, and the prediction effect is better.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.1;TP183
本文编号：2276432
[Abstract]:Short-term traffic flow prediction is a hot research field in intelligent transportation. Short-term traffic flow prediction is one of the important problems in traffic control and vehicle navigation, and is also one of the key technologies of intelligent traffic control and traffic guidance. Short-term traffic flow has the characteristics of nonlinearity, time-varying, uncertainty, instability and so on. Using various forecasting models to forecast short-term traffic flow can alleviate urban traffic congestion and avoid the waste of social resources. Therefore, the study of short-time traffic flow prediction has important practical significance and application value. For the traffic flow prediction of nonlinear time series, the RBF neural network has a good effect and its prediction accuracy is higher, but the RBF neural network has the disadvantage of poor generalization. In order to overcome the shortcoming of RBF neural network, support vector machine (SVM) is used to predict traffic flow. The data samples of traffic flow are usually thousands or so, so the error of forecasting traffic flow with support vector machine is still large. In order to overcome the shortcomings of RBF neural network and support vector machine, this paper combines the two models to achieve different combination prediction results according to different combination algorithms. Weight calculation formula is used to calculate the weights of the two prediction models, and the weight values of the two prediction results are obtained by the error of the previous prediction results, and the higher prediction results are obtained. Because this formula is based on the statistical results of empirical calculation and the prediction error is random and nonlinear, support vector machine (SVM) is used to predict the prediction results of the above two models. Support vector machine (SVM) is used to calculate the weights of two single prediction models, and a more accurate prediction result is obtained. The experimental results show that the combined model is more effective than the single model. Comparatively, the combined prediction model using support vector machine to calculate the weight value is more accurate than the combination prediction model with the formula of the right to use, and the prediction effect is better.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.1;TP183
【参考文献】
相关期刊论文 前3条
1 张敬磊;王晓原;;交通流灰色RBF网络非线性组合预测方法[J];数学的实践与认识;2011年19期
2 贺国光,马寿峰,李宇;基于小波分解与重构的交通流短时预测法[J];系统工程理论与实践;2002年09期
3 谭满春;冯荦斌;徐建闽;;基于ARIMA与人工神经网络组合模型的交通流预测[J];中国公路学报;2007年04期
，本文编号：2276432
本文链接：https://www.wllwen.com/kejilunwen/jiaotonggongchenglunwen/2276432.html
