DustStorm Prediction Challenge - Competition Rules
1. Competition Overview

This competition is based on meteorological time series data from Baotou, Inner Mongolia, aimed at predicting whether a dust storm will occur in the next 24 hours. Participants are required to use the past 72 hours of meteorological data to build a machine learning model that predicts the occurrence of a dust storm within the next 24 hours. The goal of this competition is to improve dust storm prediction accuracy to support disaster prevention, mitigation, and emergency response.

2. Participation Requirements

Submission Requirements: Participants are required to generate predictions of dust storm occurrence based on the provided test set and submit a CSV file containing the prediction results.

Prediction Format: Each row should include timestamp (timestamp) and dust_storm_prediction (prediction value, where 1 means dust storm will occur, and 0 means it will not).

The submitted CSV file must include the columns ID, timestamp, and dust_storm_prediction.

3. Dataset Description

Training Set: Contains meteorological data and dust storm occurrence labels used for training the model.

Time Range: November 15, 2025, 09:00 to November 18, 2025, 09:00.

Format: CSV file, each row contains a meteorological record and its corresponding dust storm label.

Validation Set: Used for model tuning and validation.

Time Range: November 19, 2025, 09:00 to November 20, 2025, 09:00.

Format: CSV file, containing meteorological data and dust storm labels.

Test Set: Used for evaluating model performance. The test set contains only meteorological data without labels. Participants need to generate predictions based on this data.

Time Range: November 21, 2025, 09:00 to November 22, 2025, 09:00.

Format: CSV file, containing only meteorological data for participants to make predictions.

4. Evaluation Criteria

The competition will be scored based on the following evaluation metrics:

Accuracy: The proportion of correct predictions.

AUC (Area Under the ROC Curve): The performance of the model in classification tasks.

Participants' goal is to optimize their models to improve both accuracy and AUC.

5. Submission File Format

Participants' submission files must be in CSV format and contain the following columns:

ID: A unique identifier for each row of data.

timestamp: The timestamp of the data (format: YYYY-MM-DD HH:MM:SS).

dust_storm_prediction: The dust storm prediction label (1 for occurrence, 0 for non-occurrence).

Sample File Format:

ID,timestamp,dust_storm_prediction
1,2025-11-21 09:00:00,1
2,2025-11-21 10:00:00,0
3,2025-11-21 11:00:00,1
...

6. Rules and Requirements

All participants must register and agree to the competition rules to participate.

Submitted files must adhere to the specified format.

All submissions will be automatically scored by the Kaggle system.

Final rankings will be determined based on the comparison between the predictions and the actual labels.

7. Data Usage and Licensing

The data used in this competition is for competition purposes only, and any commercial use without authorization is prohibited. All participants must comply with the rules related to data usage.

8. Awards and Rewards

Ranking awards will be based on Accuracy and AUC scores. Top-ranking participants will receive corresponding prizes and rewards.

This is the complete DustStorm Prediction Challenge competition rule set. You can further modify or add other details as needed. I hope this helps you finalize the competition rules! If you have any additional questions, feel free to let me know!