{
  "fetched_at": "2026-10-20T06:32:41Z",
  "source_url": "https://www.kaggle.com/competitions/admet-test/data",
  "pages": [
    {
      "name": "data-description",
      "content": "This page appears alongside the data files. It describes what files have been provided and the format of each. There is no single format for this page that is appropriate for all competitions, but you should strive to describe as much as you can here. A little time spent describing the data here can save a lot of time answering questions later.\n\n## Files\n\n*   **train.csv** - the training set\n*   **test.csv** - the test set\n\n\n## Columns\n\n*   `Drug_ID` - The ID of Compound\n*   `Drug` - SMILES of Drug Compound\n*   `Y` - The Activity Label (0 or 1)",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "This competition is the ADMET prediction of \"AI for Drug Discovery\" (SI364)\n\nQSAR is a widely used computational method in drug discovery and materials science that predicts properties such as biological activity, toxicity, and solubility using molecular structure descriptors and related activity values.\n\nThe theme of this competition is to train an accurate QSAR model using provided datasets, different molecular descriptors, feature selection, and machine learning algorithms to predict the properties of unknown compounds.\n\nNote that introducing additional structure-activity relationship data for the same target is not allowed in the competition to keep all participants focused on machine learning.\n\nParticipants can use common methods such as linear regression, support vector machines, or neural networks, but we encourage trying new neural network models such as Transformer and GNN.\n\nWinners will be selected based on the AUROC of participants' prediction and submitted code. Each participant can submit their code twice per day, and please upload the final code before the end of the competition.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "**Evaluation Metric**\n\nThe evaluation metric for this competition is AUC. It is a performance metric used in machine learning for evaluating the accuracy of a binary classification model. In binary classification problems, the output of the model is either 0 or 1, which corresponds to negative and positive classes, respectively. The AUC metric measures the ability of the model to distinguish between these two classes by calculating the area under the receiver operating characteristic (ROC) curve. The ROC curve plots the true positive rate (TPR) against the false positive rate (FPR) at various classification thresholds.\n\nThe AUC value ranges from 0 to 1, with higher values indicating better performance. An AUC value of 0.5 indicates that the model is no better than random guessing, while an AUC value of 1 indicates perfect classification.\nAUC is a popular metric because it is robust to class imbalance and threshold selection, which are common problems in binary classification tasks.\n\n**Submission Format**\n\nSubmission files should contain two columns: Drug_ID and Predicted_Score.\nThe file should contain a header and have the following format:\n\nDrug_ID, Predicted_Score\n\n14747200, 0.348\n24790087, 0.895",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": " - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}