Abstract: In view of missing data issue of traffic detection, this paper proposed a new short-term traffic flow composite forecasting model. The model adopted reconstruction method to solve the missing data problem, and used improved Kalman smoothing to implement short-term traffic flow forecasting. The model resolved the defeats of traditional forecasting methods which cannot deal with the missing data, and also can attain a high forecasting precision. Through the validation of Shenzhen data and being compared with the traditional methods, it has been proved that the new method has high forecasting precision, the forecasting result can maintain at 88% or more, and the model also has good practicality.

Key words: data missing, traffic flow, wavelet de-noising, Kalman smooth filtering, short-term forecasting

摘要： 针对交通检测中数据的缺失问题，提出了一种新的交通流综合短时预测模型，这种模型可以对交通检测中的缺失数据进行重建，并在此基础上运用改进的卡尔曼平滑算法进行短时交通流预测。该模型克服了传统的预测方法无法对检测数据的缺失进行处理的缺点，能在数据缺失时进行有效的交通流预测。通过深圳市的实际流量数据的验证，并比对传统方法，证实该方法具有较好的预测性能，模型预测精度可以保持在88%以上，具有较好的实用性。

关键词: 数据缺失, 交通流, 小波降噪, 卡尔曼平滑滤波, 短时预测