Predict Introverts vs. Extroverts - Kaggle Playground Series S5E7

⭕ Ongoing project - not complete

This repository contains a Google Colab notebook for the Kaggle Playground Series Season 5, Episode 7 competition, focused on predicting whether individuals are introverts or extroverts based on personality traits.

Overview

The Kaggle Playground Series S5E7 is a binary classification task where the goal is to predict the Personality (Introvert or Extrovert) using features like time spent alone, social event attendance, and stage fear. This notebook performs exploratory data analysis (EDA), data preprocessing, and builds a baseline model to establish a starting point for the competition.

Dataset

• Source: Kaggle competition dataset (train.csv, test.csv, sample_submission.csv).
• Features:
  • Numerical: Time_spent_Alone, Social_event_attendance, Going_outside, Friends_circle_size, Post_frequency.
  • Categorical: Stage_fear, Drained_after_socializing.
  • Target: Personality (Introvert/Extrovert).
• Size: 18,524 rows, 9 columns (including id).
• Challenges: Missing values (~5-10% per feature), potential class imbalance.

Notebook Contents

The notebook (s5e7_eda.ipynb) includes:

1. Data Loading: Downloads the dataset using the Kaggle API in Google Colab.
2. Exploratory Data Analysis (EDA):
   • Checks dataset shape, missing values, and feature distributions.
   • Visualizes target distribution and feature relationships using seaborn.
   • Analyzes correlations between numerical features and the target.
3. Preprocessing:
   • Imputes missing values (median for numerical, mode for categorical).
   • Encodes categorical features (Stage_fear, Drained_after_socializing) using one-hot encoding.
4. Baseline Model: Trains a logistic regression model and evaluates AUC-ROC.
5. Submission: Prepares predictions for the test set in the required Kaggle format.

Setup Instructions

To run the notebook in Google Colab:

1. Clone the Repository:

   git clone https://github.com/your-username/your-repo-name.git

2. Upload to Colab:
   • Open Google Colab.
   • Upload s5e7_eda.ipynb from the cloned repository.
3. Set Up Kaggle API:
   • Download your Kaggle API key (kaggle.json) from Kaggle Settings.
   • In Colab, run:

     from google.colab import files
     files.upload()  # Upload kaggle.json
     !mkdir -p ~/.kaggle
     !cp kaggle.json ~/.kaggle/
     !chmod 600 ~/.kaggle/kaggle.json

4. Install Dependencies:
   • The notebook installs required libraries (e.g., pandas, seaborn, scikit-learn).
   • Install the Kaggle CLI:

     !pip install kaggle

5. Download Dataset:
   • The notebook includes code to download the dataset:

     !kaggle competitions download -c playground-series-s5e7
     !unzip playground-series-s5e7.zip

Running the Notebook

• Open s5e7_eda.ipynb in Colab.
• Execute cells sequentially to load data, perform EDA, preprocess, and train the model.
• Modify the notebook to experiment with other models (e.g., XGBoost, CatBoost) or preprocessing techniques.

Results (still ongoing project)

• Baseline AUC-ROC: [Add your baseline model’s performance here, e.g., 0.75].
• Leaderboard Rank: [Add your rank or score after submission, if applicable].
• Next Steps: Experiment with feature engineering, advanced models, or hyperparameter tuning to improve performance.

Requirements

• Python 3.x (provided by Colab)
• Libraries: pandas, numpy, seaborn, matplotlib, scikit-learn, kaggle
• Kaggle API key (kaggle.json)

License

This project is licensed under the MIT License.

Acknowledgments

• Kaggle for providing the dataset and competition platform.
• Built using Google Colab for accessibility and reproducibility.

For questions or contributions, open an issue or contact me.

Made with ❤️ by scythe410.