{
  "fetched_at": "2026-10-19T11:47:47Z",
  "source_url": "https://www.kaggle.com/competitions/MLSP2023-volunteers-part2/data",
  "pages": [
    {
      "name": "data-description",
      "content": "## Files\n*   **colab_train.csv** - Training dataset, containing records of train volunteers participating in tasks during **feature time span** (2020.2-2021.8). Details of the columns can be found in the next session.\n*   **colab_train_edge_list.csv** - A graph of volunteers derived from the collaboration in the **colab_train.csv**. Nodes are volunteers, edges denote the collaboration between two volunteers, and edge weights represent the count of collaboration between two volunteers.\n*   **colab_test.csv** - Testing feature dataset, containing records of test volunteers participating in tasks during **feature time span** (2020.2-2021.8).\n*   **colab_test_candidate_multistep.csv** - This file contains all candidate links to predict. Each `Id` identifies a link, and it's the join of two volunteer IDs with an integer from 0 to 5 representing the timestamp. The 0-5 timestamp corresponds to each month in the **predict time span** (2021.8 - 2022.1).  For example, `13-141-0` means the edge between volunteer 13 and 141, at time step 0 (2021.8). `141-127469-1` means the edge between volunteer 141 and 127469 at time step 1 (2021.9).  Competitors should predict whether each edge will exist and output to a column `Predicted`, both probs or binary outputs (0,1) are acceptable.\n*   **sample_submission_collaboration_multistep.csv** - A submission sample with correct format.\n\n## Columns\n\n*   `user_id` - The unique id of a volunteer\n*   `is_organizer` -  Whether the volunteer is the organizer of the task (binary)\n*   `task_id` -The unique id of a task\n*   `task_location` - The street name of the task location (text)\n*   `task_district` - The district name of the task location (text)\n*   `user_location` - The home address of the volunteer (text)\n*   `longitude` - The longitude of the task location\n*   `latitude` - The latitude of the task location\n*   `task_start_time` - The start time of the task\n*   `task_end_time` - The end time of the task\n*   `node1` - Unique ID of volunteer A\n*   `node2` - Unique ID of volunteer B\n*   `weight` - Collaboration times between volunteer A and volunteer B",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "# Announcement💥\n## [2023.07.15] New Dates for Paper Submission\nDear Participants，here are some updates on the evaluation and paper submission process of this challenge:\n1. Challenge Winner and Paper Selection Rules:\n🚩:Winners will be selected according to the best performance for each single task, separately.  Therefore, one winner for each task will be selected.\n🚩:Up to four papers will be accepted to be presented at the conference.\n🚩:Papers will be peer reviewed as in a regular conference submission. They will be scored based on their model performance on the private test data (75%) and paper quality (25%). Paper quality metrics include clarity of method, correctness, reproducibility and originality (bonus).\n2. Paper Summary Submission:\n🚩:Submit a one-page summary about your methodology and numerical results by **July 20th**.\n🚩:Please follow the paper submission instructions on MLSP website. Make sure to choose the track **“MLSP2023 - Data Competition Papers” on CMT**.\n🚩:If you have already finished writing the paper, you can use it instead of the summary.\n🚩:You must submit the initial summary to be eligible for the final paper submission.\n3. Final Paper Submission:\n🚩:You can upload the full version of your paper to the paper submission system before the final submission deadline.\n🚩:The final paper deadline is **July 27th**.\n\n## [2023.06.21] Dataset Correction\nHello colleagues, we made some important updates regarding the **Task 2** dataset. The affected files, **`colab_train.csv`** and **`colab_train_edge_list.csv`**, have been modified to adjust the training data time span to a more realistic setting. The new time range is **\"2020.2-2021.8\"** instead of \"2020.2-2022.2\".  \n \nPlease make sure to use the training dataset with the **corrected time span** when training your model. Furthermore, when submitting your paper, please **include your code**. \n\nPlease note that these changes **may affect the leaderboard**. We sincerely apologize for the inconvenience this may cause and appreciate your cooperation.  \n\n# About this Task👀\nRecent years have seen a rise in online crowdsourcing platforms that facilitate volunteering efforts, allowing individuals to contribute to collective goals and address community needs. To optimize these efforts, it is crucial to understand volunteer behavior and collaboration patterns. In this challenge, we present a volunteer participation dataset with two tasks:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F9176866%2Fd902a44e3d8e4097d23f39211feb7e24%2Fslide8(1).gif?generation=1687097913843193&alt=media)\n\n**Volunteer Dataset**: The dataset is collected from a mobile crowdsourcing app called “Anti-Pandemic Pioneer” used for organizing volunteers during the COVID-19 pandemics in Shenzhen, China. It contains records of each volunteer participating in a task, including features like volunteer ID, task ID, timestamp, task location, etc. Task names and descriptions will also be provided.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F9176866%2F95e49f9816dd5fd63af1987ab3297717%2Fmap.png?generation=1687256611473169&alt=media)\n\n**Task 2: Future Collaboration Prediction:** Our objective is to predict potential collaborations between pairs of volunteers who will continue their volunteering efforts. As shown in the purple highlighted areas in the above figure, we have selected two districts, Futian District and Longhua District, from the data as the dataset for this task. The available resources consist of volunteer activities spanning from February 2020 to July 2021, along with an edge list serving as the ground truth. Our goal is to forecast the collaboration between pairs of volunteers on a monthly basis over a 6-month period, ranging from August 2021 to February 2022. Predicting future collaborations holds diverse applications, including gaining insights into group dynamics within volunteer communities and developing social recommendations for volunteers.\n\n**Key Points & Competition Details:**\n- Participants can choose to participate in either one or both tasks.\n- The datasets for both tasks have identical structure but sampled from distinct districts, ensuring no overlap.\n- The evaluation metrics are Root Mean Squared Error(RMSE) for Task 1 and Area Under Receiver Operating Characteristic curve (AUC-ROC) for Task 2.\n- Both datasets are split into a train set and a test set. The test sets are further split into private test set (2/3 for final evaluation) and public test set (1/3 for getting feedbacks during the competition).\n- Participants can access all datasets but only labels for the train sets.\n- Submissions can be made for the public test set twice a day, and immediate scores (RMSE and AUC-ROC) will be provided.\n- Submission deadline will be one months after the start of the competition.\n\n**If you have any questions or inquiries regarding this competition, please do not hesitate to contact us. You can reach out to us through [our Slack channel] (https://join.slack.com/t/volunteer-challenge/shared_invite/zt-1xq09saep-50EqZDlw7ds5EHN3Lbp1_g).**\n\n**To access further details about the competition, please visit our [competition website](https://volunteerchallenge.github.io/).**\n\n**Please NOTE.** that the intellectual property (IP) is not transferred to the challenge organizers, i.e., if code is shared/submitted, the participants remain the owners of their code (when the code is made publicly available, an appropriate license should be added).\n\n**Privacy Notice.** The volunteer activity data used in this challenge were passively collected from the mobile platform with personal identification information removed. No investigations on humans were involved. During the preparation of this challenge, the organizers did not have access to any identification information associated with the collected data. We are committed to adhering to strict privacy and ethical guidelines during the entire competition process.\n\n**Acknowledgement.** We thank the Shenzhen Pioneers APP operation team for providing the volunteer data.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "**Please NOTE.** that the intellectual property (IP) is not transferred to the challenge organizers, i.e., if code is shared/submitted, the participants remain the owners of their code (when the code is made publicly available, an appropriate license should be added).\n\n**Privacy Notice.** The volunteer activity data used in this challenge was passively collected from the mobile platform with personal identification information removed. No investigations on humans were involved. During the preparation of this challenge, the organizers did not have access to any identification information associated with the collected data. We are committed to adhering to strict privacy and ethical guidelines during the entire competition process.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "The goal of this task is to forecast the collaboration between pairs of volunteers on a monthly basis over a 6-month period, ranging from August 2021 to February 2022.\n\n## Submission Format\n\n- The file should contain two columns named `Id` and `Predicted`.\n- The `Id` column should consist of three hyphen-connected elements. These elements represent pairwise volunteers and a timestep `i` ranging from 0 to 5, indicating the prediction for the next `i-th` month.\n- Each `Predicted` value in the file represents the probability of collaboration between the corresponding two participants.\n\nThe file should contain a header and have the following format. Also, a sample submission form is given at [Data](https://www.kaggle.com/competitions/MLSP2023-volunteers-02/data?select=sample_submission_collaboration_multistep.csv) Page. \n```\nId,Predicted\n122028-204227-0,0.7\n185807-155353-3,0.3\n185807-155353-5,0.4\n```\n\n## Area Under Receiver Operating Characteristic Curve (AUC-ROC Curve)\n\nThe evaluation metric for this competition is the [Area Under Receiver Operating Characteristic Curve (AUC-ROC Curve)](https://towardsdatascience.com/understanding-auc-roc-curve-68b2303cc9c5). The AUC-ROC curve is a graphical representation of the performance of a binary classification model. It plots the True Positive Rate (TPR) against the False Positive Rate (FPR) for different classification thresholds. The area under this curve, known as AUC-ROC, is commonly used as a metric to evaluate the model's performance.\n\n### Formula\n\nThe TPR and FPR are calculated using the following formulas:\n\nTPR = TP / (TP + FN)\n\nFPR = FP / (FP + TN)\n\nWhere:\n- TP (True Positives): The number of positive instances correctly classified as positive.\n- FN (False Negatives): The number of positive instances incorrectly classified as negative.\n- FP (False Positives): The number of negative instances incorrectly classified as positive.\n- TN (True Negatives): The number of negative instances correctly classified as negative.\n\n### Properties\n\n- The AUC-ROC value ranges between 0 and 1, where a higher value indicates better performance. An AUC-ROC of 0.5 suggests a random classifier, while an AUC-ROC of 1 indicates a perfect classifier.\n- The curve is generated by varying the classification threshold and calculating the corresponding TPR and FPR values.\n- The AUC-ROC curve is commonly used in machine learning to compare different models or to optimize classification thresholds.\n- AUC-ROC is insensitive to class imbalance, making it a suitable metric when dealing with imbalanced datasets.\n- The AUC-ROC curve is a useful tool for visualizing the trade-off between sensitivity and specificity in a classification model.\n\n### Interpretation\n\nBy examining the AUC-ROC curve, you can determine how well your model is performing across different classification thresholds. The curve's shape can provide insights into the model's behavior and its ability to discriminate between positive and negative instances.\n\nA model with a curve closer to the top-left corner indicates higher overall accuracy, while a curve closer to the diagonal line suggests poor performance.\n\nRemember that the AUC-ROC curve is just one tool for evaluating model performance, and it should be used in conjunction with other metrics and domain knowledge to make informed decisions.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Timeline",
      "content": "Important Dates:\n- Jul 19, 2023: Deadline for Submitting Results\n- Jul 20, 2023: Paper Summary Submission\n- Jul 27, 2023: Final Paper Submission\n- Aug 15, 2023: Notification of paper acceptance\n- Aug 25, 2023: Camera-Ready Challenge Papers Deadline",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Announcement",
      "content": "## [2023.07.15] New Dates for Paper Submission\nDear Participants，here are some updates on the evaluation and paper submission process of this challenge:\n1. Challenge Winner and Paper Selection Rules:\n🚩:Winners will be selected according to the best performance for each single task, separately.  Therefore, one winner for each task will be selected.\n🚩:Up to four papers will be accepted to be presented at the conference.\n🚩:Papers will be peer reviewed as in a regular conference submission. They will be scored based on their model performance on the private test data (75%) and paper quality (25%). Paper quality metrics include clarity of method, correctness, reproducibility and originality (bonus).\n2. Paper Summary Submission:\n🚩:Submit a one-page summary about your methodology and numerical results by **July 20th**.\n🚩:Please follow the paper submission instructions on MLSP website. Make sure to choose the track **“MLSP2023 - Data Competition Papers” on CMT**.\n🚩:If you have already finished writing the paper, you can use it instead of the summary.\n🚩:You must submit the initial summary to be eligible for the final paper submission.\n3. Final Paper Submission:\n🚩:You can upload the full version of your paper to the paper submission system before the final submission deadline.\n🚩:The final paper deadline is **July 27th**.\n\n\n## [2023.06.21] Dataset Correction\nHello colleagues, we made some important updates regarding the **Task 2** dataset. The affected files, **`colab_train.csv`** and **`colab_train_edge_list.csv`**, have been modified to adjust the training data time span to a more realistic setting. The new time range is **\"2020.2-2021.8\"** instead of \"2020.2-2022.2\".  \n \nPlease make sure to use the training dataset with the **corrected time span** when training your model. Furthermore, when submitting your paper, please **include your code**. \n\nPlease note that these changes **may affect the leaderboard**. We sincerely apologize for the inconvenience this may cause and appreciate your cooperation.  ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}