Turin Micromobility Analysis

Comprehensive Spatial-Temporal Analysis of E-Scooter Sharing Services

Politecnico di Torino | Department of Regional and Urban Studies and Planning | 2024-2025

Overview | Key Findings | Data | Installation | Usage | Citation

---

Overview

This project presents a comprehensive data-driven investigation of shared micro-mobility patterns in Turin, Italy. We analyze 2,548,650 e-scooter trips across three major operators (Lime, Bird, Voi) to understand service availability, pricing structures, and the relationship between e-scooters and public transport.

Research Objectives

1. Service Level Analysis: Compare fleet density, availability, and battery levels across operators
2. Temporal Patterns: Identify peak usage hours, weekly cycles, and seasonal trends
3. Public Transport Integration: Determine if e-scooters compete with or complement public transit
4. Economic Viability: Evaluate operator profitability and cost structures
5. Optimal Mode Choice: Recommend the best operator for specific commuting scenarios

Central Research Question

"Are e-scooters competitors or allies to public transport in Turin?"

---

Key Findings

Summary Statistics

Metric	Value	Interpretation
Total Trips Analyzed	2,548,650	After data cleaning (87-99% retention)
Integration Index (200m)	99.7%	Near-universal proximity to PT stops
Feeder Rate	95.1%	Strong first/last-mile connector role
Peak Hours	8-9 AM, 17-19 PM	Clear commuting patterns
Market Size	€8.30M/year	Combined operator revenue
Probability of Loss	0.28%	Very low financial risk (Monte Carlo)

Operator Comparison

Operator	Total Trips	Revenue (€/year)	Profit Margin	Market Share
LIME	1,421,374	4,245,099	55.7%	55.8%
BIRD	852,751	3,217,369	61.5%	33.5%
VOI	274,525	837,898	53.5%	10.8%

Main Conclusion

E-scooters function primarily as first/last-mile connectors rather than direct competitors to public transport. Over 95% of trips start or end within 200 meters of a transit stop.

---

Methodology

Study Area

• City: Turin, Italy (Metropolitan Area)
• Operators: Lime, Bird, Voi
• Data Period: 12 months (2023-2024)
• Commute Route: Codegone to Castello del Valentino (3.6 km)

Snapshot Analysis Approach

Our methodology follows a systematic snapshot analysis framework:

┌─────────────────────────────────────────────────────────────────┐
│  1. DEFINE STUDY AREA                                           │
│     └── Select zone where all operators are active (~1 km²)     │
├─────────────────────────────────────────────────────────────────┤
│  2. DATA COLLECTION                                             │
│     └── Capture vehicle locations, battery, pricing from apps   │
├─────────────────────────────────────────────────────────────────┤
│  3. CALCULATE KPIs                                              │
│     └── Fleet density, market share, battery levels, costs      │
├─────────────────────────────────────────────────────────────────┤
│  4. APPLY TO COMMUTE                                            │
│     └── Calculate trip costs and generalized cost per operator  │
└─────────────────────────────────────────────────────────────────┘

Key Performance Indicators

KPI	Formula	Purpose
Fleet Density	Vehicles / Area (km²)	Measures ease of finding a scooter
Market Share	Operator Vehicles / Total Vehicles	Shows operator dominance
Average Battery	Mean(Battery Readings)	Indicates fleet quality
Trip Cost	Unlock Fee + (Duration × Rate)	Determines value for money
Integration Index	Trips near PT / Total Trips	PT complementarity

Statistical Methods

• Kruskal-Wallis H-test: Compare distributions across operators
• Chi-square Test: Categorical variable independence
• Mann-Whitney U Test: Pairwise non-parametric comparison
• Monte Carlo Simulation: 10,000 iterations for risk analysis
• Bootstrap Confidence Intervals: 95% CI (n=1,000 resamples)
• Kaplan-Meier Survival Analysis: Parking duration modeling
• Weibull Distribution Fitting: Parametric survival model
• Log-rank Test: Survival curve comparison
• Moran's I: Spatial autocorrelation detection
• STL Decomposition: Seasonal-trend analysis

---

Project Structure

Turin-MicromobilityAnalysis/
│
├── run_pipeline.py              # Master pipeline controller
├── requirements.txt             # Python dependencies
├── README.md                    # This file
├── ARCHITECTURE.md              # Technical documentation
│
├── src/
│   ├── analysis/                # Core analysis modules
│   │   ├── 01_temporal_analysis.py      # Hourly/daily/monthly patterns
│   │   ├── 02_od_matrix_analysis.py     # Origin-destination flows
│   │   ├── 03_integration_analysis.py   # E-scooter & PT comparison
│   │   ├── 04_parking_analysis.py       # Parking duration & turnover
│   │   └── 05_economic_analysis.py      # Revenue & profitability
│   │
│   ├── data/
│   │   └── 00_data_cleaning.py          # Data preprocessing
│   │
│   ├── utils/                   # Helper functions
│   │
│   └── visualization/           # Plotting scripts
│       ├── 01_temporal_dashboard.py
│       ├── 02_od_spatial_flows.py
│       ├── 03_integration_maps.py
│       ├── 04_parking_maps.py
│       └── 05_economic_sensitivity.py
│
├── outputs/
│   ├── figures/                 # Generated visualizations (PNG)
│   │   ├── exercise1/           # Temporal pattern charts
│   │   ├── exercise2/           # OD flow maps
│   │   ├── exercise3/           # Integration analysis maps
│   │   ├── exercise4/           # Parking analysis charts
│   │   └── exercise5/           # Economic analysis charts
│   │
│   ├── reports/                 # CSV and pickle checkpoints
│   │
│   └── tables/                  # Statistical summary tables
│       ├── exercise1/           # Descriptive statistics
│       ├── exercise2/           # OD matrices
│       ├── exercise3/           # Integration metrics
│       ├── exercise4/           # Parking statistics
│       └── exercise5/           # Financial summaries
│
└── docs/                        # Additional documentation

---

Data Download

The complete dataset (~6.8 GB) is hosted on Kaggle.

Option 1: Kaggle CLI (Recommended)

# Install Kaggle CLI
pip install kaggle

# Download dataset (requires Kaggle API key)
kaggle datasets download -d aliivaezii/turin-escooter-trips -p data/ --unzip

Option 2: Manual Download

1. Visit kaggle.com/datasets/aliivaezii/turin-escooter-trips
2. Click Download
3. Extract to data/ folder

Dataset Contents

Processed Data (4.4 GB):

File	Size	Records
lime_cleaned.csv	2.0 GB	1.4M
bird_cleaned.csv	106 MB	850K
voi_cleaned.csv	53 MB	275K
df_all.pkl	2.2 GB	2.5M

Raw Data (2.4 GB):

Folder	Size	Contents
lime/	2.2 GB	CSV + 16 monthly files
bird/	86 MB	2024 + 2025 CSV
voi/	28 MB	22 Excel files
gtfs/	70 MB	PT network data
zone_statistiche_geo/	344 KB	Shapefile

---

Analysis Pipeline

Exercise 1: Temporal Pattern Analysis

Analyzes hourly, daily, and monthly usage patterns for each operator.

Methods:

• Seasonal-Trend decomposition using Loess (STL)
• Peak detection using scipy signal processing
• Statistical comparison with Kruskal-Wallis test

Key Outputs:

• Hourly trip distribution with confidence intervals
• Weekly usage heatmaps
• Monthly trend analysis
• Fleet utilization rates by operator

Exercise 2: Origin-Destination Matrix

Maps mobility corridors and zone-to-zone flows across Turin.

Methods:

• Spatial joining to official Turin Zone Statistiche
• Advanced OD metrics (Gini, Entropy, Flow Asymmetry)
• Chi-square tests for temporal independence
• Hierarchical clustering for zone grouping

Key Outputs:

• OD flow matrices (zone-to-zone)
• Spatial flow maps with arrow thickness
• Trip density hexbin maps
• Top corridor identification

Exercise 3: Public Transport Integration

Evaluates e-scooter proximity to public transport stops.

Methods:

• Multi-buffer sensitivity analysis (50m, 100m, 200m)
• Moran's I for spatial autocorrelation
• Chi-square tests for integration independence
• Permutation tests (1,000 iterations)

Key Outputs:

• Integration index by buffer distance
• Competition map (PT-parallel routes)
• Feeder service identification
• Temporal comparison with PT schedules

Exercise 4: Parking Analysis

Studies parking duration patterns and turnover rates.

Methods:

• Kaplan-Meier survival curves with 95% CI
• Weibull distribution fitting (MLE)
• Log-rank test for group comparisons
• Ghost vehicle detection (>120 hours)

Key Outputs:

• Parking duration histograms
• Survival curves by operator
• Parking intensity heatmaps
• Turnover vs. demand scatter plots

Exercise 5: Economic Analysis

Calculates revenue, fleet economics, and profitability.

Methods:

• Monte Carlo simulation (10,000 iterations)
• Break-even analysis with sensitivity testing
• Pareto analysis (80/20 rule for zones)
• Scenario modeling (base/optimistic/pessimistic)

Key Outputs:

• Operator P&L waterfall charts
• Revenue yield maps
• Break-even scatter plots
• Sensitivity tornado diagrams

---

Installation

Prerequisites

• Python 3.10 or higher
• pip package manager
• Git

Quick Start

# Clone the repository
git clone https://github.com/aliivaezii/Turin-MicromobilityAnalysis.git
cd Turin-MicromobilityAnalysis

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # macOS/Linux
# .venv\Scripts\activate   # Windows

# Install dependencies
pip install -r requirements.txt

Dependencies

Package	Version	Purpose
pandas	≥2.0.0	Data manipulation
numpy	≥1.24.0	Numerical computing
scipy	≥1.10.0	Statistical analysis
geopandas	≥0.14.0	Geospatial analysis
shapely	≥2.0.0	Geometric operations
matplotlib	≥3.7.0	Visualization
seaborn	≥0.12.0	Statistical graphics
contextily	≥1.3.0	Basemap tiles
tqdm	≥4.65.0	Progress bars

---

Usage

Run Complete Pipeline

python run_pipeline.py

Run Specific Exercises

# Run only exercises 1, 2, and 3
python run_pipeline.py --stages 1 2 3

# Run from exercise 3 onwards
python run_pipeline.py --from-stage 3

# Run only visualizations (skip analysis)
python run_pipeline.py --viz-only

# Run analysis without visualizations
python run_pipeline.py --no-viz

Pipeline Stages

Stage	Module	Description	Runtime
0	Data Cleaning	Clean and harmonize raw data	~2 min
1	Temporal Analysis	Hourly/daily/monthly patterns	~5 min
2	OD Matrix	Origin-destination flow analysis	~8 min
3	Integration	E-scooter and PT comparison	~10 min
4	Parking	Duration and turnover analysis	~5 min
5	Economic	Revenue and profitability	~15 min

---

Results

Sample Visualizations

Temporal Patterns (Exercise 1)

Public Transport Integration (Exercise 3)

Economic Analysis (Exercise 5)

---

Cost Comparison for Commuters

For a typical Home-to-University commute (Codegone to Valentino, 18 minutes):

Operator	Unlock Fee	Rate/min	Trip Cost	Monthly (44 trips)
Lime	€1.00	€0.25	€5.50	€242.00
Bird	€1.00	€0.19	€4.42	€194.48
Voi	€1.00	€0.19	€4.42	€194.48

Subscription Recommendations

Scenario	Best Choice	Reason
Occasional use (< 10 trips/month)	Any operator	Similar pricing
Regular commute (20+ trips/month)	Lime Pro subscription	Fixed monthly fee
Short trips (< 5 min)	Bird or Voi	Lower per-minute rate
High availability priority	Lime	Largest fleet

---

Technical Notes

Data Quality

• Raw data cleaned using multi-stage pipeline
• Geographic filtering to Turin metropolitan boundary
• Temporal filtering for valid trip durations (1-60 minutes)
• Battery level validation (0-100%)

Statistical Rigor

• All confidence intervals at 95% level
• Effect sizes reported (Cohen's d, eta-squared)
• Non-parametric tests used for non-normal distributions
• Bootstrap resampling for robust estimates

---

Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/new-analysis)
3. Commit changes (git commit -m 'Add new analysis module')
4. Push to branch (git push origin feature/new-analysis)
5. Open a Pull Request

---

Author

Ali Vaezi

Interuniversity Department of Regional and Urban Studies and Planning
Politecnico di Torino
Corso Duca degli Abruzzi, 24, Turin 10129, Italy

Email: ali.vaezi@studenti.polito.it

---

Citation

If you use this code or dataset in your research, please cite:

@misc{ali_vaezi_2026,
	title={Turin Escooter Trips 2024},
	url={https://www.kaggle.com/dsv/14486854},
	DOI={10.34740/KAGGLE/DSV/14486854},
	publisher={Kaggle},
	author={Ali Vaezi},
	year={2026}
}

DOI: 10.34740/KAGGLE/DSV/14486854

---

License

This project is licensed under the MIT License. See LICENSE for details.