Abstract: With consideration of the economy development tendency in China, a civil vehicle population prediction model is developed based on the space-time theory. First, with the domestic and international experience on vehicle development, the gray Verhulst model is used to describe the vehicle population development tendency and predict the vehicle population in the next 20 years. Second, several social and economical indexes related to vehicle population are selected by comparing the correlation coefficient value, and then the principal component method is used to reduce dimension of the selected indexes and obtain some principal indexes. Based on the econometrics theory, a forecasting model is formulated to predict the vehicle population in the next 20 years. Integrating these two forecasting models, a non-liner combination forecasting model is developed based on the BP neural network. The reliability and accuracy of the linear combination forecasting model are tested by the vehicle data from 2003 to 2007. Finally, the civil vehicle population of China in the next 20 years is predicted again based on the linear combination forecasting model.

Key words: urban traffic, non-liner prediction, space-time theory, grey Verhulst model, neural network

摘要： 基于时空理论，结合未来我国经济发展趋势，对中国民用汽车拥有量进行预测. 首先，根据国内外汽车发展经验，利用灰色Verhulst模型拟合中国汽车拥有量的发展规律，并初步估计中国未来20年汽车拥有量；然后，根据关联理论筛选出与民用汽车拥有量关系密切的若干个社会经济指标，运用主成分分析法对这些指标进行降维，提取出几个主要的彼此不相关的综合指标，运用计量经济学理论建立民用汽车拥有量预测模型，并再次估计中国未来20年汽车拥有量；结合上述两个模型的优点，利用BP神经网络建立非线性组合预测模型. 2003～2007年数据证实该非线性组合预测模型具有更强的可靠性和预测精度. 最后利用该非线性组合预测模型对中国未来20年的民用汽车拥有量进行预测.

关键词: 城市交通;非线性预测;时空论;灰色Verhulst模型, 神经网络