Results of the 2nd Traffic Jam Prediction Challenge Contest
-Challenging the sophistication of congestion prediction on Expressway using expressway toll and route search data-
- Corporate Top
- Press Room
- Press Release Headquarters
- Results of the 2nd Traffic Jam Prediction Challenge Contest
2024年7月31日
東日本高速道路株式会社
東京大学大学院情報学環
NEXCO東日本（東京都千代田区）と東京大学大学院情報学環（東京都文京区）は、渋滞予測モデルの精度を競う『第2回渋滞予測チャレンジコンテスト（以下、コンテスト）』を2024年1月31日に発表し、最終審査及び表彰式を6月27日に開催しました。
コンテストでは、211件の応募があり、この中から精度賞10名、モデリング賞2名の入賞者を決定しました。
(center left/center: S_O, center right: tomoe)
1. Challenge Contest Results
The top 10 winners of the Accuracy Award were selected based on the data provided by our company (traffic volume and congestion data, route search data, etc.) and the models that had the smallest discrepancy between predictions and actual results for the specified dates between April 1 and May 6, 2024. The top three winners were awarded prizes. In addition, two winners who developed the best congestion prediction models were awarded the Modeling Award after qualitatively evaluating the data processing methods and modeling policies of the winners.
Accuracy Award
| User ID | Prediction accuracy * | |
|---|---|---|
| First place | tomoe | 0.56422 |
| 2nd place | S_O | 0.54637 |
| 3rd place | mameg | 0.54557 |
- Prediction accuracy is an index that considers both recall and precision, and takes a value between 0 and 1. The higher the value, the higher the accuracy.
- Recall rate: The percentage of traffic jams that actually occurred that could be predicted to occur
- Precise rate: The ratio of traffic jams that actually occurred to those predicted to occur
- For information on other winners, please visit the competition website "SIGNATE".
modeling award
| User ID | Reason for commendation |
|---|---|
| tomoe | The models were divided by route and direction, and features were selected and set to reflect the sequence of days of the week during peak traffic periods, and the patterns and locations of congestion, making the model close to the thinking of NEXCO EAST 's congestion forecasters. |
| S_O | The model considered the main causes of traffic congestion to be "temporal factors" such as mornings, evenings, and consecutive holidays, and "geographical factors" such as the proximity of tourist spots and In-bound, and was designed with ingenuity in mind, such as setting as features the proportion of ordinary cars in DraPla route search data and groups classified based on congestion trends from the traffic counter. |
Judge's comment (Professor, Interfaculty Initiative in Information Studies, University of Tokyo, Noboru Koshizuka)
Following on from the first contest, we asked the contestants to develop excellent congestion prediction models this time as well. Although we changed the data provided from the first contest, we received many ideas from new perspectives, such as data processing methods and modeling policies.
Comment from the Secretariat (Takeshi Kawasaki, General Manager of the ITS Promotion Department, NEXCO EAST Management Business Headquarters)
We will confirm the validity of the features and ideas of the traffic congestion prediction model proposed this time, and then work towards realizing our "mo Vision for the next generation Expressway."
2. Future plans
The winning congestion prediction model will be analyzed and efforts will be made to further improve congestion prediction, such as by applying it to other peak traffic periods, such as the Obon and New Year holidays, and expanding it to other routes. In addition, through joint research, the two companies will consider automating data cleansing and linking it with in-house systems in order to commercialize the model, as well as providing highly accurate and timely congestion prediction information.
3. judge
Professor Noboru Koshizuka, Graduate School of Interdisciplinary Information Studies, The University of Tokyo
Associate Professor, Graduate School of Interdisciplinary Information Studies, The University of Tokyo
Mobility journalist Etsuko Kusuda
NEXCO Research Institute, Traffic Environment Research Department, Director of Traffic Research Xing Jian Xin Jiang
Nobuyoshi Nakanishi, General Manager NEXCO EAST Management Business Headquarters
To view the PDF file, you need the Adobe Systems plug-in software "Acrobat Reader (Japanese version)". If you do not have Download from here (free)Please use it.