Imputing Vehicle Detector Data by Genetic Programming

• Main Authors: Shih-lun Chen, 陳世倫
• Other Authors: Jiann-Sheng Wu
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/29986039998215015343

碩士 === 國立中央大學 === 土木工程研究所 === 100 === To search for the optimal imputation of Vehicle Detector, this paper, we carried out an empirical analysis for missing value of Hshehshan tunnel via Genetic Programming. We, at first, use signal attribute data impute missing value by accumulated nearest pairs of up- and downstream vehicle detectors, and analyze the relation between performance and number of vehicle detectors, whereupon, we imputed missing value by multi-attribute data. After testing data imputation, we ranked all types of imputation according to the performance. Finally, Recurrent Neural Network was selected to compare with Genetic Programming. The results showed that the performance of Genetic programming is better than Recurrent Neural Network. If we ranked all types of imputation according to conbined the three imputation performance, the rank as follows, flow&speed imputaiotn is 1st, speed&flow&occ imputation is 2st and speed imputation is the worst. For flow imputation, we use flow data, the accumulated nearest five pairs of detector up- and downstream could be input for the highest accuracy. For speed imputation, we use flow&speed data, the accumulated nearest three pairs of detector up- and downstream could be input for the highest accuracy. For occupancy imputation, we use occupancy data, the accumulated nearest eleven pairs of detector up- and downstream could be input for the highest accuracy.