{ "fetched_at": "2026-10-19T11:05:57Z", "source_url": "https://www.kaggle.com/competitions/BT5153-2025-whales-prediction-challenge/data", "pages": [ { "name": "data-description", "content": "Competitors are provided with two rich datasets:\n\n**User Profile Data**: This dataset encapsulates a variety of user-specific attributes to help model financial behavior, including:\n - `user_id`: A unique identifier for each user.\n - `mobile_brand_name`, `mobile_marketing_name`: Information on the user's mobile device.\n - `age_in_year`, `gender_name`, `marital_status`, `education_background`, `income_level`, `occupation`: Demographic and socio-economic details of the user.\n\n**Transaction Data** (April 1, 2022, to May 31, 2022): This dataset details users' financial activities, providing a lens into their transaction habits through attributes such as:\n - `asset_type`, `transaction_time`, `transaction_type`, `GTV`: Specifics of each transaction.\n - `user_id`: Links back to the user profile for integrated analysis.\n\nThe challenge is to use this data to forecast the transaction volume category (\"high\" or \"low\") of users in June 2022, thereby identifying potential \"whales.\"\n\nTraining data is provided in `train_label.csv`. It has two attributes: `user_id` and `tgt` that indicates users'tag: 1 indicates Whales while 0 is Normal Users\n\n**Output Data**:\n- Submissions must be in a CSV format with two columns: `user_id` and `pred_prob`, where:\n - `user_id` is the unique identifier for each user.\n - `pred_prob` is the model's prediction of the likelihood (between 0 and 1) that the user is a \"whale.\"\n- Participants can submit up to 3 entries per day.\n- **Example Submission Format**: you can check `submission_sample.csv`\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Objective", "content": "The goal is to build a model that not only predicts high vs. low transaction volumes with high accuracy but also sheds light on the underlying factors that contribute to higher financial engagement among users. This insight is invaluable for businesses aiming to tailor their services, optimize marketing strategies, and enhance customer engagement by identifying and understanding their most valuable customers.\n\n**Mandatory GenAI/LLM Usage**\nAs part of this kaggle competition, everyone MUST leverage Large Language Models (LLMs) or other Generative AI techniques throughout their competition. This requirement aims to: \n- Familiarize participants with real-world AI tool integration \n- Explore innovative ways to combine traditional ML with modern AI tools \n- Develop best practices for AI-assisted data science\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Business Impacts", "content": "Predicting \"whales\" has far-reaching implications across various business functions:\n\n- **Marketing and Sales**: Knowing which users are likely to transact more can help in crafting personalized marketing campaigns, optimizing resource allocation, and improving sales strategies to maximize ROI.\n \n- **Customer Experience**: Insights into the transaction behavior can guide the development of tailored products and services, enhancing the user experience and fostering loyalty among high-value customers.\n\n- **Risk Management**: Understanding the transaction patterns of users can aid in assessing financial risks and fraud detection, leading to more robust risk management strategies.\n\n- **Strategic Planning**: Businesses can use predictions to inform their strategic decisions, focusing on areas with the highest potential for growth and investment.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "**One account per participant**\n- You may not sign up for this Kaggle competition using multiple accounts and therefore may not make submissions from multiple accounts.\n\n**No private sharing outside teams**\n- You may not privately share code or data with data with another or other participants.\n- You also should not upload your code in public platforms such as Github.\n\n**No plagiarism**\n- There is a zero-tolerance policy regarding plagiarism.\n\n**No usage of external data**\n- You may not use any data other than the training and test data provided. Your submission must be reproducible only with the code in your notebook to be submitted and the training and test data. Any attempt to hack this competition will have the consequence of a score penalty.\n\n**EXCEPTION**: You may use pre-trained feature extraction pipelines, like GPT but are suggested not to train any of those large models.\n\n**Submission Limits**\n- You may submit a maximum of 3 entries per day.\n- You may select up to 2 final submissions for judging.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "abstract", "content": "The competition is designed for those passionate about leveraging machine learning to make significant predictions in financial behavior. This challenge offers a unique opportunity to predict users' future transaction volumes—classifying them as either \"high\" or \"low\" compared to their peers in the subsequent month. By analyzing a comprehensive dataset of user profiles and their recent transaction history, competitors will unlock key insights and identify patterns that influence transactional behavior.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Description", "content": "# **BT5153-2025-Whales Prediction Challenge**\n\n## **Important**\n- Upon signing up for Kaggle and joining the competition, kindly change the Display Name in your Kaggle profile and your Team Name under the Team tab to your NUS student number, e.g. **\"A0123456Z\"**.\n- Do not sign up using more than one Kaggle account.\n- This is a competition between individual participants.\n- Please read the **Overview, Submission Guidelines, Grading, Data**, and **Rules** pages carefully.\n- Please note with respect to the **End Time** of this competition (see Timeline) that it is specified in UTC.\n\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "Entries will be evaluated on the **Area Under the ROC Curve (AUC) metric**, emphasizing the model's precision in distinguishing between the two classes of transaction volumes.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Submission Guidelines", "content": "## Important Dates:\n- Kaggle Competition Release Date: 14th March 2025\n- Kaggle Competition Deadline: 4th April 2025\n- GenAI Integration Paper Deadline: 18th April 2025\n\n## Submission Format on Canvas\n**1. GenAI Integration Paper**: \n- Name your paper with your NUS Student Number, i.e. **A0123456Z.pdf**. \n- Length: 2-3 pages (excluding references)\n- Format: PDF\n- Use the following templates from ACL: https://2023.aclweb.org/calls/style_and_formatting/\n- Include tables and graphs in an appendix (2 pages limit)\n- Citations for AI tools and resources used\n\n**2. Jupyter Notebook**: \n- Submit a Jupyter notebook (.ipynb) for the final model for your top submission, together with your paper. Name this file with your NUS Student Number, i.e. **A0123456Z.ipynb**.\n- Note: Your notebook should be in working condition. That means after reading your report the grader might try to run all cells of the notebook. If the notebook fails to run it might attract some penalty. Thus don’t leave dangling code in the notebook and test the notebook yourself once before submitting.\n\n##Required GenAI Integration Paper Submission\n**1. AI Tool Selection and Strategy**\n- Description of chosen AI tools (e.g., ChatGPT, Claude, GitHub Copilot, Cursor, Devin)\n- Rationale for tool selection\n- Integration strategy with traditional ML pipeline\n\n**2. Implementation and Usage**\n- Detailed examples of AI tool usage in the competition (not limited to the following):\n - Data preprocessing and feature engineering\n - Model selection and architecture design\n - Code optimization and debugging\n - Result analysis and interpretation\n- Document prompt engineering strategies\n- Show iterations and refinements in AI interactions\n- Demonstrate understanding of AI tool capabilities and limitations\n- Here, participants are encouraged to document specific cases where AI tools failed or hindered development, including incorrect suggestions, increased complexity, or debugging difficulties and provide in-depth analysis.\n - Analysis should include root causes of AI failures and practical recommendations for when to avoid AI tool usage in similar scenarios.\n\n**3. Impact Analysis**\n- Quantitative/Qualitative assessment of AI tool impact on:\n - Development time and efficiency\n - Model performance\n - Code quality\n- Critical evaluation of AI tool limitations\n- Lessons learned and best practices identified\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Grading", "content": "This competition will constitute 20% of your total points for this module. This 20% will be divided into two sections: Model Performance and GenAI Integration Paper.\n\n## 1. Model Performance (5 points)\nYour grading will be based on your AUC scores in the private leaderboard. If your score is larger than the baseline score that we have demonstrated, you will get the full points here.\n- Beat the baseline model: 5 points\n- Below baseline: 0 points\n\n## 2. GenAI Integration Paper (15 points)\n\nYour report **must include** the following 3 sections describing:\n\n- Innovation in AI tool usage (5 points)\n- Implementation quality and depth (5 points)\n- Critical analysis and insights (5 points)\n\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }