Gym User Dropout Prediction Dataset
- kaggle.com
Updated Aug 3, 2025
Share
Facebook
Twitter
EmailClick to copy linkLink copiedCiteHassan Abdul-razeq (2025). Gym User Dropout Prediction Dataset [Dataset]. https://www.kaggle.com/datasets/hassanabdulrazeq/gym-user-dropout-prediction-datasetCroissantCroissant is a format for machine-learning datasets. Learn more about this at mlcommons.org/croissant.Dataset updatedAug 3, 2025AuthorsHassan Abdul-razeqLicensehttps://creativecommons.org/publicdomain/zero/1.0/https://creativecommons.org/publicdomain/zero/1.0/
DescriptionGym User Dropout Prediction Dataset
Description
This synthetic dataset simulates user behavior in a fitness application, designed to predict the risk of gym membership dropout based on attendance patterns and personal attributes. The dataset contains 10,000 realistic user profiles with features that influence gym retention, making it ideal for classification tasks in behavioral analytics.
Key Features
- Realistic distributions matching actual gym user behavior patterns
- Complex feature interactions that simulate real-world decision-making
- Controlled noise to mimic natural data variability
- Balanced classes for effective machine learning modeling
Potential Use Cases
- Predicting at-risk users for retention interventions
- Analyzing factors contributing to gym commitment
- Developing personalized workout recommendations
- Behavioral segmentation of fitness app users
Dataset Characteristics
- Number of instances: 10,000
- Number of features: 8 predictive + 1 target
- Missing values: No
- Synthetic but realistic: Yes
Columns Description
Feature Type Description Value Range user_id int Unique user identifier 1-10000 age int User's age 18-60 (peaked at 25-40) gender categorical User's gender Male/Female sessions_per_week int Weekly gym attendance 0-7 sessions avg_session_duration float Average workout length in minutes 10-120 progress_score float Composite fitness progress metric 0-100 mood_after categorical Post-workout emotional state Energized/Neutral/Fatigued injury categorical Reported workout injuries None/Knee/Back/Shoulder dropout binary Target variable - quit status 0 (active)/1 (quit) Generation Methodology
Data was programmatically generated with: 1. Base distributions matching real gym statistics 2. Logical correlations between features (e.g., more sessions → longer durations) 3. Non-linear relationships in target variable 4. Controlled noise injection (Gaussian + categorical variability)
Suggested Evaluation Metrics
For classification models: - Precision-Recall curves (class imbalance consideration) - F1 score - ROC AUC - Feature importance analysis
License
CC0: Public Domain (Free to use for any purpose)
Acknowledgements
Synthetic dataset created for machine learning education and benchmarking purposes. Inspired by real fitness app analytics challenges.
Dataset Link
Not seeing a result you expected?
Learn how you can add new datasets to our index.
FacebookTwitterEmailhttps://creativecommons.org/publicdomain/zero/1.0/https://creativecommons.org/publicdomain/zero/1.0/
Gym User Dropout Prediction Dataset
Description
This synthetic dataset simulates user behavior in a fitness application, designed to predict the risk of gym membership dropout based on attendance patterns and personal attributes. The dataset contains 10,000 realistic user profiles with features that influence gym retention, making it ideal for classification tasks in behavioral analytics.
Key Features
- Realistic distributions matching actual gym user behavior patterns
- Complex feature interactions that simulate real-world decision-making
- Controlled noise to mimic natural data variability
- Balanced classes for effective machine learning modeling
Potential Use Cases
- Predicting at-risk users for retention interventions
- Analyzing factors contributing to gym commitment
- Developing personalized workout recommendations
- Behavioral segmentation of fitness app users
Dataset Characteristics
- Number of instances: 10,000
- Number of features: 8 predictive + 1 target
- Missing values: No
- Synthetic but realistic: Yes
Columns Description
| Feature | Type | Description | Value Range |
|---|---|---|---|
| user_id | int | Unique user identifier | 1-10000 |
| age | int | User's age | 18-60 (peaked at 25-40) |
| gender | categorical | User's gender | Male/Female |
| sessions_per_week | int | Weekly gym attendance | 0-7 sessions |
| avg_session_duration | float | Average workout length in minutes | 10-120 |
| progress_score | float | Composite fitness progress metric | 0-100 |
| mood_after | categorical | Post-workout emotional state | Energized/Neutral/Fatigued |
| injury | categorical | Reported workout injuries | None/Knee/Back/Shoulder |
| dropout | binary | Target variable - quit status | 0 (active)/1 (quit) |
Generation Methodology
Data was programmatically generated with: 1. Base distributions matching real gym statistics 2. Logical correlations between features (e.g., more sessions → longer durations) 3. Non-linear relationships in target variable 4. Controlled noise injection (Gaussian + categorical variability)
Suggested Evaluation Metrics
For classification models: - Precision-Recall curves (class imbalance consideration) - F1 score - ROC AUC - Feature importance analysis
License
CC0: Public Domain (Free to use for any purpose)
Acknowledgements
Synthetic dataset created for machine learning education and benchmarking purposes. Inspired by real fitness app analytics challenges.