🚴 Tour de France Data Analysis & Dashboard

A comprehensive Business Intelligence project analyzing historical Tour de France data from 1903-2020, featuring interactive dashboards, anomaly detection, and predictive modeling.

📊 Project Overview

This project transforms over a century of Tour de France historical data into actionable insights through:

• Interactive Power BI Dashboard
• Machine Learning Models (Anomaly Detection & Victory Prediction)
• Data Warehouse Design
• Comprehensive ETL Pipeline

🎯 Business Objectives

• Identify performance factors over time (distance, duration, speed)
• Analyze winner distribution by country
• Study participation trends (number of finishers)
• Track evolution of total distance and average time
• Create a historical dashboard for sports analysts and media

📁 Project Structure

France_tour/ ├── Documents/ │ └── Project Presentation.pdf ├── EDA/ │ └── Exploratory Data Analysis notebooks ├── ML algorithms/ │ ├── anomaly_detection.py │ └── victory_prediction.py ├── PowerBI Dashboard and webservice/ │ └── Tour_de_France_Dashboard.pbix └── .gitattributes

🛠️ Technologies Used

• Data Analysis: Python (Pandas, NumPy)
• Machine Learning: Scikit-learn (Isolation Forest, Random Forest)
• Data Visualization: Power BI
• Data Processing: ETL with Python
• Version Control: Git/GitHub

📈 Key Features

📊 Dashboard KPIs

• Overall Metrics: Editions count, participants, winner speed/time, nations diversity
• Country Analysis: Wins, riders, stage victories, finishers
• Rider Analysis: Performance metrics, podium rates, competitiveness
• Team Analysis: Victory rates, stage wins, team performance
• Stage Analysis: Distance by type, stage composition, winner dominance

🤖 Machine Learning Models

1. Anomaly Detection

• Algorithm: Isolation Forest
• Purpose: Identify unusual Tour editions
• Features: Distance, stages count, starters, finishers
• Results: Detected 5 anomalous editions (1903-1905, 1919, 2020)

2. Victory Prediction

• Algorithm: Random Forest
• Purpose: Predict stage winners
• Performance: AUC 0.85, Precision 82%
• Key Insight: Distance is the most important factor

🗃️ Data Warehouse Schema

Dimensions:

• Dim Stage (stage details, type, locations)
• Dim Winner (rider statistics, performance metrics)
• Dim Tour (edition overview, dates, participants)
• Dim Finisher (rider rankings, country, team)

🔧 Data Processing

Datasets Cleaned:

• df_finishers (9,895 rows) - Time normalization, missing value handling
• df_stages (2,341 rows) - Location separation, distance normalization
• df_tours (109 rows) - Date separation, distance processing
• df_winners (102 rows) - Time correction, unit standardization

ETL Challenges Solved:

• Complex time format conversion (94h 33' 14" → seconds)
• Geographical data separation
• Missing value imputation
• Unit standardization

🌐 Web Accessibility

• Published Dashboard: Power BI Online Service
• Features: Interactive visualizations, dynamic filters, multi-page navigation
• Access: Secure web link for authorized users

👥 Target Users

• Race Organizers (ASO): Event evolution visualization
• Sports Journalists: Enriched analytical content
• Sponsors: Country-specific visibility analysis
• Cycling Fans: Historical trend exploration

🚀 Getting Started

Prerequisites

• Python 3.8+
• Power BI Desktop
• Required Python packages: pandas, scikit-learn, numpy

Installation

git clone https://github.com/Hajer5503/France_tour.git
cd France_tour
pip install -r requirements.txt