CSIE 5043 Machine Learning Final Project

[2023/12/02 Update] new station: 500109089

[2023/11/26 Update] new station: 500107143

[2023/11/11 Update] new station: 500110084

[2023/11/08 Update] new station: 500103065

Description

The repository is used to daily release training data for the regular track final project to the learners of the NTU course.

• Useful Links:
  • Course Website: from the lecturer's homepage
  • Final Project Guildlines: https://www.csie.ntu.edu.tw/~htlin/course/ml23fall/final/final.pdf
  • Kaggle competitions:
    • First Stage: https://www.kaggle.com/competitions/html2023fall-final
      • public test period: 2023/10/21 00:00 ~ 2023/10/24 23:40; private test period: 2023/12/04 ~ 2023/12/10; learners must submit predictions before 12/3 23:59
    • Second Stage: https://www.kaggle.com/competitions/html2023fall-final-project-stage-2
      • test period: 2023/12/11 00:00 ~ 2023/12/17 23:40; learners must submit predictions before 12/10 23:59
    • Third Stage (tentative): https://www.kaggle.com/competitions/html2023fall-final-project-stage-3
      • test period: 2023/12/18 00:00 ~ 2023/12/24 23:40; learners must submit predictions before 12/17 23:59

*All timestamp are in UTC+8

Metadata

• demographic.json: file contains demographic information for each station in the following format:

  {
      sno: the station ID number for each bike station in string
          {
              "sna": the name of the station in Traditional Chinese,
              "sarea": the district of the station in Traditional Chinese,
              "lat": the latitude of the station in float,
              "lng": the longtitude of the station in float,
              "ar": the address of the station in Traditinal Chinese,
              "sareaen": the district of the station in English,
              "aren": the address of the staion in English,
          }
  ...
  }
  

• sno_test_set.txt: contains the stations' sno that needs to predict in the test set
• release/$date: directory for all training data, where the $date is in YYYYMMDD format
  • in each directory contains ${sno}.json files, which record the data of each station with number $sno of the coresponding date in the following format:

  {
      timestamp: string in 24-hour format (HH:MM)
          {
              "tot": maximum capacity number of the station at the minute, integers > 0
              "sbi": the current bike amount of the station at the minute, integers >= 0
              "bemp": the number of slots available for parking bikes at the minute, integers >= 0
              "act": whether the station is open to use at the minute, where "1" means open and "0" means prohibited, strings
          }
      ...
  }
  

  • we will update release/ on a daily base.
• sample_submission_stage${i}.csv: sample submission for each stage in csv format, e.g.

  id,sbi
  20231021_500101001_06:40,7.89
  ...
  

  The first row of your submission must be id,sbi, the following rows must contain 2 columns and seperated by ,. The first column is the id of your prediction in YYYYMMDD_${sno}_HH:MM, where ${sno} must be one of the station number appears in sno_test_set.txt, YYYYMMDD indicates the format of date, and HH:MM is the format of time. (Y=year, M=month, D=day, H=hour, M=minute). The second column is your bike amount prediction of the bike station at the given timestamp, it can be integer or float. The number of rows should be equal to 88705, that is, 1 row of id,sbi and 88704 rows of predictions (112 stations $\times$ 3 predictions per hour $\times$ 24 hours $\times$ (4 days in public set period + 7 days in private set period).

For learners that are not familiar with JSON format, please refer to https://en.wikipedia.org/wiki/JSON.

How to use

You can use the following commands to get the data, where the latest will be released on a daily frequency.

$ git clone git@github.com:hyusterr/html.2023.final.data.git

# use this command to get the latest released data after 13:00 every day
$ git pull

If you want to be notified through your GitHub e-mail whenever there is a update in this repository, please click Watch on the up-right corner.

For learners who are new to git and Github, please refer to the tutorials made by the GitHub offical to setting up the environment for yourself.

• Quickstart: https://docs.github.com/en/get-started/quickstart
• Generating a new SSH key and adding it to the ssh-agent: https://docs.github.com/en/authentication/connecting-to-github-with-ssh/generating-a-new-ssh-key-and-adding-it-to-the-ssh-agent