Upload README.md (Parquet format for HF viewer)

README.md

@@ -1,304 +1,138 @@
 ---
 license: mit
 task_categories:
-  - time-series-forecasting
-language:
-  - en
+- time-series-forecasting
 tags:
-  - heart-rate-prediction
-  - running
-  - sports-science
-  - physiological-modeling
-  - endomondo
+- heart-rate
+- running
+- physiological-modeling
+- lstm
+- endomondo
 size_categories:
-  - 10K<n<100K
+- 10K<n<100K
 ---
 
-# Heart Rate Prediction Dataset (Endomondo V2)
+# Endomondo Heart Rate Prediction Dataset V2
 
-Clean, preprocessed dataset for predicting heart rate from running workout data (speed and altitude).
+## Dataset Summary
 
-## Dataset Description
+This dataset contains **40,186 running workouts** from **761** athletes, designed for heart rate prediction from speed and altitude time-series.
 
-This dataset contains **40,186 running workouts** from the Endomondo fitness tracking platform, quality-filtered and smoothed for heart rate prediction using machine learning.
+Each workout includes:
+- **Time-series**: Heart rate (target), speed, altitude, timestamps
+- **Metadata**: Workout type, duration, user ID
+- **Statistics**: Pre-computed HR/speed metrics for filtering
 
-### Dataset Statistics
+## Dataset Structure
 
-- **Total Workouts**: 40,186
-- **Total Users**: 761 unique runners
-- **Sport**: Running only
-- **Format**: JSON (single file with all workouts)
-- **File Size**: 1.4 GB
+### Splits
 
-### Workout Types Distribution
+| Split | Workouts | Description |
+|-------|----------|-------------|
+| Train | 28,130 | Training set (70%) |
+| Validation | 6,027 | Validation set (15%) |
+| Test | 6,029 | Test set (15%) |
 
-| Type | Count | Description |
-|------|-------|-------------|
-| RECOVERY | 15,095 | Easy pace runs (HR mean < 120 BPM) |
-| STEADY | 22,991 | Constant moderate pace |
-| INTENSIVE | 2,100 | High intensity workouts |
-
-## Data Format
-
-The dataset is a single JSON file with the following structure:
-
-```json
-{
-  "metadata": {
-    "timestamp": "2026-01-14T15:07:10",
-    "original_count": 46250,
-    "final_count": 40186,
-    "removed": {
-      "flagged_samples": 5830,
-      "total_removed": 6064
-    },
-    "workout_type_counts": {...},
-    "flags_applied": [...],
-    "smoothing_inherited": {...}
-  },
-  "workouts": [
-    {
-      "workout_id": 296982347,
-      "user_id": 4969375,
-      "sport": "run",
-      "workout_type": "RECOVERY|STEADY|INTENSIVE",
-      "duration_min": 108.38,
-      "data_points": 500,
-      "heart_rate": [103.0, 105.2, ...],  // BPM values
-      "speed": [8.89, 9.12, ...],          // km/h
-      "altitude": [34.85, 35.2, ...],      // meters
-      "timestamp": [1392480163, ...],      // Unix timestamps
-      "hr_mean": 134.9,
-      "hr_std": 10.3,
-      "hr_min": 75,
-      "hr_max": 163,
-      "speed_source": "GPS_computed",
-      "speed_metrics": {...},
-      "workout_type_onehot": {
-        "RECOVERY": 1,
-        "STEADY": 0,
-        "INTENSIVE": 0
-      }
-    },
-    ...
-  ]
-}
-```
-
-### Fields Description
-
-**Workout Metadata**:
-- `workout_id`: Unique identifier from Endomondo
-- `user_id`: Anonymized user ID
-- `sport`: Always "run" in this dataset
-- `workout_type`: Categorized as RECOVERY, STEADY, or INTENSIVE
-- `duration_min`: Total workout duration in minutes
-- `data_points`: Number of timesteps (typically 500, ~6 seconds per point)
-
-**Time Series Data** (arrays of equal length):
-- `heart_rate`: Heart rate in BPM
-- `speed`: Running speed in km/h (computed from GPS when missing)
-- `altitude`: Elevation in meters
-- `timestamp`: Unix timestamps
-
-**Statistics**:
-- `hr_mean`, `hr_std`, `hr_min`, `hr_max`: Heart rate statistics
-- `speed_metrics`: Speed data quality and statistics
-- `workout_type_onehot`: One-hot encoding for workout type
-
-## Data Quality & Preprocessing
+### Features
 
-### Quality Filtering
+| Feature | Type | Description |
+|---------|------|-------------|
+| `workout_id` | int | Unique workout identifier |
+| `user_id` | int | Anonymous user identifier |
+| `workout_type` | string | RECOVERY, STEADY, or INTENSIVE |
+| `duration_min` | float | Workout duration in minutes |
+| `data_points` | int | Number of timesteps (max 500) |
+| `heart_rate` | list[float] | Heart rate time-series [BPM] |
+| `speed` | list[float] | Speed time-series [km/h] |
+| `altitude` | list[float] | Altitude time-series [meters] |
+| `timestamp` | list[float] | Unix timestamps [seconds] |
+| `hr_mean` | float | Average heart rate [BPM] |
+| `hr_std` | float | HR standard deviation |
+| `hr_min` | float | Minimum HR [BPM] |
+| `hr_max` | float | Maximum HR [BPM] |
+| `speed_mean` | float | Average speed [km/h] |
+| `speed_max` | float | Maximum speed [km/h] |
+| `altitude_gain` | float | Cumulative elevation gain [m] |
+| `split` | string | train / validation / test |
 
-Workouts were removed if they had:
-- Mean HR < 50 BPM or > 200 BPM (unrealistic)
-- HR std < 5 BPM (too constant, likely sensor error)
-- Negative HR-speed correlation < -0.3 (physiologically impossible)
-- Excessive missing data points
+### Workout Type Distribution
 
-**Total removed**: 6,064 workouts (13.1% of original dataset)
-
-### Preprocessing Pipeline
-
-1. **Speed Computation**:
-   - Calculated from GPS coordinates using haversine distance
-   - Filled missing speed values from original dataset
-   - Outliers above 25 km/h flagged and smoothed
-
-2. **Smoothing**:
-   - Applied 7-point moving average to heart_rate, speed, and altitude
-   - Reduces sensor noise while preserving workout patterns
-   - Edge cases handled with reflection padding
-
-3. **Workout Type Classification**:
-   - **RECOVERY**: HR mean < 120 BPM
-   - **INTENSIVE**: HR mean > 165 BPM OR HR max > 200 BPM
-   - **STEADY**: All others (moderate, consistent effort)
-
-## Usage
-
-### Loading the Dataset
-
-```python
-import json
-from huggingface_hub import hf_hub_download
-
-# Download file
-dataset_path = hf_hub_download(
-    repo_id="rricc22/endomondo-hr-prediction-v2",
-    filename="clean_dataset_v2.json",
-    repo_type="dataset"
-)
-
-# Load data
-with open(dataset_path, 'r') as f:
-    data = json.load(f)
-
-# Access workouts
-workouts = data['workouts']
-metadata = data['metadata']
-
-print(f"Total workouts: {len(workouts)}")
-print(f"First workout ID: {workouts[0]['workout_id']}")
-```
+| Type | Count | Description |
+|------|-------|-------------|
+| RECOVERY | 15,095 | Easy runs (low intensity) |
+| STEADY | 22,991 | Moderate pace runs |
+| INTENSIVE | 2,100 | High intensity workouts |
 
-### Example: Filter by Workout Type
+## Data Quality
 
-```python
-# Get all intensive workouts
-intensive = [w for w in workouts if w['workout_type'] == 'INTENSIVE']
+All workouts have been:
+1. **Filtered** for quality (removed HR anomalies, corrupted data)
+2. **Smoothed** with 7-point moving average (reduces GPS noise)
+3. **Validated** against physiological constraints:
+   - HR mean ≥ 120 BPM
+   - HR max ≤ 200 BPM
+   - HR std ≥ 5 BPM
+   - Speed-HR correlation ≥ -0.3
 
-print(f"Intensive workouts: {len(intensive)}")
-```
+Removed: 6,064 low-quality workouts
 
-### Example: Train/Val/Test Split
+## Usage Example
 
 ```python
-import numpy as np
-
-# Get unique users
-users = list(set(w['user_id'] for w in workouts))
-np.random.seed(42)
-np.random.shuffle(users)
+from datasets import load_dataset
 
-# Split users 70/15/15
-n_train = int(0.7 * len(users))
-n_val = int(0.15 * len(users))
+# Load full dataset
+dataset = load_dataset("rricc22/endomondo-hr-prediction-v2")
 
-train_users = set(users[:n_train])
-val_users = set(users[n_train:n_train+n_val])
-test_users = set(users[n_train+n_val:])
+# Access splits
+train_data = dataset['train']
+test_data = dataset['test']
 
-# Split workouts by user
-train_workouts = [w for w in workouts if w['user_id'] in train_users]
-val_workouts = [w for w in workouts if w['user_id'] in val_users]
-test_workouts = [w for w in workouts if w['user_id'] in test_users]
+# Example workout
+workout = train_data[0]
+print(f"Workout Type: {workout['workout_type']}")
+print(f"Duration: {workout['duration_min']:.1f} min")
+print(f"Avg HR: {workout['hr_mean']:.1f} BPM")
+print(f"Avg Speed: {workout['speed_mean']:.1f} km/h")
 
-print(f"Train: {len(train_workouts)} workouts from {len(train_users)} users")
-print(f"Val:   {len(val_workouts)} workouts from {len(val_users)} users")
-print(f"Test:  {len(test_workouts)} workouts from {len(test_users)} users")
-```
-
-### Example: Extract Features for ML
-
-```python
-import numpy as np
-
-def extract_features(workout):
-    """Extract input features and target for a single workout."""
-    # Input features
-    speed = np.array(workout['speed'])
-    altitude = np.array(workout['altitude'])
-    gender = 1 if workout.get('gender', 'Male') == 'Male' else 0
-    
-    # Target
-    heart_rate = np.array(workout['heart_rate'])
-    
-    return {
-        'speed': speed,
-        'altitude': altitude,
-        'gender': gender,
-        'heart_rate': heart_rate,
-        'length': workout['data_points']
-    }
-
-# Extract features for all workouts
-features = [extract_features(w) for w in workouts]
+# Access time-series
+heart_rate = workout['heart_rate']  # List of HR values
+speed = workout['speed']            # List of speed values
 ```
 
 ## Model Performance
 
-A 2-layer LSTM model trained on this dataset achieved:
-- **MAE**: 7.42 BPM on test set
-- **RMSE**: 9.54 BPM
-
-**Trained Model**: [heart-rate-prediction-lstm](https://huggingface.co/rricc22/heart-rate-prediction-lstm)  
-**Interactive Demo**: [Heart Rate Predictor](https://huggingface.co/spaces/rricc22/heart-rate-predictor)
+This dataset was used to train an LSTM model achieving:
+- **7.42 BPM** Mean Absolute Error
+- **17% improvement** over baseline
 
-## Intended Use
+See the model card: [rricc22/heart-rate-prediction-lstm](https://huggingface.co/rricc22/heart-rate-prediction-lstm)
 
-### Primary Use Cases
-- Training machine learning models for heart rate prediction
-- Research on physiological response modeling during exercise
-- Time-series forecasting benchmarking
-- Sports science and exercise physiology studies
+Try the demo: [Heart Rate Predictor](https://huggingface.co/spaces/rricc22/heart-rate-predictor)
 
-### Out-of-Scope
-- Medical diagnosis or treatment decisions
-- Non-running activities (cycling, swimming, etc.)
-- Real-time monitoring applications
+## Source
 
-## Limitations
-
-1. **Population**: Primarily European recreational runners from Endomondo platform (2014-2016 era)
-2. **Device Variability**: Mixed GPS accuracy and heart rate sensor quality
-3. **Speed Range**: Most data in 8-15 km/h range (marathon training pace)
-4. **Missing Demographics**: Only user_id available, no age/weight/fitness level
-5. **Environmental Factors**: No temperature, humidity, wind, or terrain type data
-
-## Ethical Considerations
-
-- **Privacy**: Original Endomondo public dataset (users consented to data sharing)
-- **Anonymization**: No personally identifiable information (user IDs anonymized)
-- **Not Medical**: For research and training optimization only, not medical use
-- **Bias**: Model trained on this data may not generalize to all populations or fitness levels
+- **Original Data**: Endomondo dataset
+- **Processing Pipeline**: Quality filtering → Smoothing → Feature engineering
+- **Version**: V2 (January 2026)
+- **License**: MIT
 
 ## Citation
 
 If you use this dataset, please cite:
 
 ```bibtex
-@dataset{endomondo_hr_v2_2026,
-  author = {Riccardo},
-  title = {Heart Rate Prediction Dataset (Endomondo V2)},
-  year = {2026},
-  publisher = {Hugging Face},
-  url = {https://huggingface.co/datasets/rricc22/endomondo-hr-prediction-v2}
-}
-```
-
-**Original Endomondo Dataset**:
-```bibtex
-@article{endomondo2016,
-  title={The Endomondo dataset},
-  author={Gjoreski, Martin and others},
-  journal={Available online},
-  year={2016}
+@dataset{endomondo_hr_v2,
+  title={Endomondo Heart Rate Prediction Dataset V2},
+  author={Riccardo},
+  year={2026},
+  publisher={Hugging Face},
+  url={https://huggingface.co/datasets/rricc22/endomondo-hr-prediction-v2}
 }
 ```
 
-## License
-
-MIT License - See LICENSE file for details
-
 ## Related Resources
 
-- **Model**: [heart-rate-prediction-lstm](https://huggingface.co/rricc22/heart-rate-prediction-lstm) - 2-layer LSTM (7.42 BPM MAE)
-- **Demo**: [Heart Rate Predictor](https://huggingface.co/spaces/rricc22/heart-rate-predictor) - Interactive Streamlit app
-- **Original Dataset**: Endomondo HR public dataset
-
----
-
-**Created**: January 14, 2026  
-**Version**: 2.0  
-**Contact**: For questions or issues, please open an issue on the dataset repository.
+- 🤗 **Model**: [heart-rate-prediction-lstm](https://huggingface.co/rricc22/heart-rate-prediction-lstm)
+- 🚀 **Demo**: [Interactive Predictor](https://huggingface.co/spaces/rricc22/heart-rate-predictor)
+- 📊 **GitHub**: [SUB3_V2 Repository](https://github.com/rricc22/SUB3_V2)