| DC 欄位 |
值 |
語言 |
| DC.contributor | 土木工程學系 | zh_TW |
| DC.creator | 張騰文 | zh_TW |
| DC.creator | Teng-wen Chang | en_US |
| dc.date.accessioned | 2012-8-23T07:39:07Z | |
| dc.date.available | 2012-8-23T07:39:07Z | |
| dc.date.issued | 2012 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=993202076 | |
| dc.contributor.department | 土木工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 由於國內外文獻尚未利用基因規劃法(Genetic Programming，GP)預測高速公路旅行時間，且近年來國道高速公路局將電子收費系統(Electronic Toll Collection，ETC)所偵測之交通資料開放索取，又高速公路主要交通資料來自車輛偵測器(Vehicle detector，VD)，故本研究期望透過車輛偵測器(VD)及電子收費(ETC)所偵測之交通資料，利用基因規劃法預測高速公路旅行時間，提供精準之旅行時間預測，以作為用路人路徑選擇或是出發時間決策判斷之依據。
本研究研究範圍為國道三號樹林收費站至龍潭收費站，以此範圍之車輛偵測器資料及電子收費資料加以分析，透過偵測器資料、電子收費資料與基因規劃法，建立高速公路旅行時間預測模式，故本研究主要目的為處理車輛偵測器資料以作為輸入資料、篩選研究範圍之電子收費資料進而求得電子收費旅行時間、車輛偵測器資料與電子收費資料之整合、以輸入變數和輸入偵測器之方式利用基因規劃法建立旅行時間預測模式、以電子收費之旅行時間資料驗證旅行時間預測模式。最後，將基因規劃法之預測績效與國內外文獻績效進行比較。
結果顯示，利用基因規劃法預測高速公路旅行時間可獲得優良之預測績效，其平均績效介於4.87%~10.04%之間，而且與國內外旅行時間文獻之績效相當。綜合輸入變數預測而言，以速度預測之績效最佳，其次為流量&速度，而流量&占有率績效最差。其中利用VD間隔取之速度資料，可獲得最佳預測績效，平均MAPE值為4.87%。綜合輸入VD預測而言，利用VD隨機選取資料預測可獲得最佳績效，而VD全取與VD間隔取並列第二。其中利用速度之VD隨機選取資料，可獲得最佳預測績效，平均MAPE值為4.98%。
| zh_TW |
| dc.description.abstract | Owing to the literature is unused by Genetic Programming to forecast travel time on freeway all over the world ,as well as the Taiwan Area National Freeway Bureau makes Electronic Toll Collection open the traffic information from for free and freeway traffic information major from vehicle detectors. Our objective in this report is using the information of VD and ETC to provide data exactly for the choice of passerby or the base of departure time.
The study was designed to establish the range from Shulin tollbooth No. 3 to Longtan tollbooth to analyze data thronging VD, ETC and GP to establish the method of time forecast on freeway. The aim of this study was to use the data of VD which is inputted and selected by ETC data from the range of study to receive ETC travel time,VD data,Data compilation,using the way which is inputted data of VD to establish the method of travel time by GP ,and verification by ETC. Then, we get the forecast by GP to compare to literatures at home and abroad.
These results suggest that using GP to forecast travel time on freeway can get a superior consequent. The average value between 4.87 to 10.04% is just similar to other literatures. Therefore, data of VD gets the first place for speed prediction, second place is flow & speed, and flow &occ is the worst of the three. In sum, speed data of VD interval selection that the average MAPE riches 4.87% is the best prediction. From what has been discussed with forecasting on VD inputting, we can get the best performance by randomly selected, and whole selection is tied with interval selection for second place. In conclusion, random selection is the greatest performance on VD that we can get 4.98% on MAPE.
| en_US |
| DC.subject | 電子收費 | zh_TW |
| DC.subject | 基因規劃法 | zh_TW |
| DC.subject | 旅行時間預測 | zh_TW |
| DC.subject | 車輛偵測器 | zh_TW |
| DC.subject | Genetic Programming | en_US |
| DC.subject | Travel time forecasting | en_US |
| DC.subject | Vehicle detector | en_US |
| DC.subject | Electronic Toll Collection | en_US |
| DC.title | 利用基因規劃法預測高速公路旅行時間 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Forecasting Travel Time on Freeway based on Genetic Programming | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |