Kaggle Solutions Repository

Collection of different Kaggle challenges. python to use : $env:KAGGLE_CONFIG_DIR='C:\Users\avig.kaggle'; C:\Users\avig\anaconda3\python.exe submit_kaggle.py

AI Agent Instructions

When adding new challenges or solutions, you MUST follow these strict naming conventions:

1. Prefix Naming: Never use generic names like data, solutions, or utils. Always prefix them with the project/challenge name.

   • Bad: data/, solution.py
   • Good: Titanic_data/, Titanic_solution.py, Spaceship_Titanic_data/

2. Structure:

   • Kaggel_Solutions/<Challenge_Name>/<Version_Folder>/
   • Inside the version folder:
     • <Challenge_Name>_data/ (Directory for datasets)
     • <Challenge_Name>_solution.py (Main script)
     • <Challenge_Name>_download_data.py (Download script)

3. Execution:

   • Scripts must rely on relative paths or the prefixed folder names, not hardcoded "data".

4. Global Configuration Isolation:

   • Do NOT place project-wide configurations (e.g., kaggle.json, submission tokens, or generic submission scripts) inside specific solution folders.
   • These must remain in the project root or use the User's global environment (e.g., ~/.kaggle/).

Maintain this consistency to prevents conflicts and ambiguity in a multi-project environment.

Usage & Recovery (User Guide)

In case of power failure or a fresh environment start, follow these steps to resume work or run solutions. use conda

1. Python Environment Setup

To run the advanced solutions (Stacking, Boosting), you need the following packages installed.

Recommended Installation Command:

pip install pandas numpy scikit-learn kaggle xgboost lightgbm catboost

Key Libraries:

• Core: pandas, numpy (Data Manipluation)
• Models: scikit-learn (RandomForest, Linear), xgboost, lightgbm, catboost (Gradient Boosting SOTA)
• API: kaggle (Data download/submission)

2. Kaggle API Setup (Required for Submission)

To automate downloading data and submitting predictions, we use a local kaggle.json.

Kaggle Submission Commands

Titanic:

kaggle competitions submit -c titanic -f submission.csv -m "Titanic Submission"

House Prices:

kaggle competitions submit -c house-prices-advanced-regression-techniques -f submission.csv -m "House Prices Submission"

Spaceship Titanic:

kaggle competitions submit -c spaceship-titanic -f submission.csv -m "Spaceship Titanic Submission"

3. Security:
   • No Hardcoded Keys: The kaggle.json file is ignored by git to prevent accidental exposure of credentials.
   • Setup:
     1. Rename kaggle.json.example to kaggle.json.
     2. Add your Kaggle username and key to the file.
     3. Alternatively, place your kaggle.json in ~/.kaggle/ (default system location).
   • File Content (kaggle.json.example):

   {
     "username": "<KEY>",
     "key": "<KEY>"
   }

   • The .gitignore in this project is configured to exclude kaggle.json.

4. Automated Submission Workflow (Reliable Method)

Since kaggle.exe path issues are common on Windows, a dedicated Python script is included to handle submission using the local credentials.

Option A: Universal Submission Script (Recommended)

1. Open manual_submit.py and ensure the SUBMISSION_FILE path points to your desired CSV.
2. Run:

   python manual_submit.py

Option B: CLI Direct Command If you prefer the command line, use python -m kaggle to avoid path issues:

# Set Config Path (PowerShell)
$env:KAGGLE_CONFIG_DIR = "c:\Users\avig\Documents\GitHub\Kaggel_chalenge_1"

# Submit
python -m kaggle competitions submit -c <competition-name> -f <path-to-csv> -m "Your Message"

4. Project Structure Overview

• Kaggel_Solutions/: Root for all challenges.
  • <Challenge>/: (e.g., Titanic, Spaceship_Titanic)
    • <Version>/: (e.g., v1_baseline, 03_Stacking_SOTA)
      • *_solution.py: The logic.
      • *_download_data.py: Data fetcher.

5. Backward Compatibility & Isolation

• Each version folder (e.g., v1_baseline, v2_advanced) is self-contained.
• They may share the *_data folder definition, but they contain their own logic and download scripts if necessary.
• Do not overwrite older versions. Always create a new version folder (e.g., v3_experimental) for new attempts.