Here is the English translation of your project description:


Project Context

Airline ticket prices depend on many factors: flight duration, booking date, number of stopovers, airline, etc.

You work for a booking platform and must:

  • Understand the factors influencing prices
  • Analyze the data
  • Build a reliable predictive model

The dataset contains 300,000+ flight bookings.

👉 The goal is to predict the price with controlled error.

🎯 Objectives

  • Perform a rigorous exploratory data analysis (EDA)
  • Identify relationships between variables and the target
  • Formulate and test statistical hypotheses
  • Build a reliable preprocessing pipeline
  • Train and compare several regression models
  • Evaluate a final model

👉 Business objective: MAE < €15 with confidence interval

📊 Project Phases

1️⃣ Data Loading

  • Import necessary libraries
  • Load the dataset Clean_Dataset.csv
  • Check the shape and structure of the data

2️⃣ Technical Variable Analysis

  • Objective: understand the nature of each variable
  • Identify discrete and continuous variables
  • Check data types
  • Identify necessary transformations

3️⃣ Univariate Analysis

  • Objective: analyze each variable independently
  • For each variable:
    • Categorical variables: class distribution, visualization (bar plot)
    • Numerical variables: descriptive statistics, histogram, and boxplot

4️⃣ Multivariate Analysis

  • Objective: understand relationships with price
  • Study the impact of:
    • Flight duration vs. price
    • Airline vs. price
    • Number of stopovers vs. price
    • Days before flight vs. price
  • Use: boxplots, binning, visual analyses

5️⃣ Hypothesis Formulation and Testing

  • Objective: statistically validate observations
  • Formulate H0 / H1 hypotheses
  • Choose the appropriate statistical test:
    • Correlation (Pearson)
    • ANOVA
  • Interpret the p-value (5% significance level)

6️⃣ Preprocessing

  • Objective: prepare data for the model
  • Handle missing values
  • Encode categorical variables
  • Transform ordinal variables
  • Standardize numerical variables
  • Build an sklearn pipeline

7️⃣ Modeling

  • Objective: compare multiple models
  • Test: Baseline model (DummyRegressor), Linear Regression, Ridge Regression, and Random Forest
  • For each model:
    • Cross-validation
    • MAE / RMSE / R²
    • Performance comparison

8️⃣ Final Evaluation

  • Objective: select the best model
  • Choose the final model
  • Train it on the full training data
  • Evaluate on the test set

9️⃣ Confidence Interval

  • Objective: quantify uncertainty
  • Compute absolute errors
  • Estimate MAE
  • Calculate standard error
  • Build a 95% confidence interval

🔟 Model Export

  • Objective: save the final model
  • Save the complete pipeline (preprocessing + model)
  • Save model metadata
  • Prepare the model for future use

👥 Pedagogical Modalities

Organization

  • Group work
  • Mandatory collaboration across all project stages

⏳ Duration

  • 5 working days: April 20 → April 24, 2026, before 5:00 PM

🧰 Tools Used

  • Python
  • Pandas / NumPy (data manipulation and analysis)
  • Seaborn / Matplotlib (visualization)
  • Scikit-learn (modeling and ML pipeline)

Evaluation Modalities

1️⃣ Oral Presentation (10 minutes)

  • Exploratory Data Analysis (EDA)
  • Main business and statistical insights
  • Final model selected and justification

2️⃣ Practical Demonstration (10–15 minutes)

  • Complete preprocessing pipeline
  • Comparison of several regression models
  • Justification of the final model choice

3️⃣ Technical Questions (5–10 minutes)

  • Statistical tests used and their interpretation
  • Preprocessing steps (encoding, scaling, etc.)
  • Model choices and comparison logic

Deliverables

  • 📓 A well-structured and readable Jupyter notebook
  • 📊 Relevant and interpreted visualizations
  • 💾 A final model exported in .joblib format
  • 🗂️ A clean GitHub repository with README
  • 📽️ A PowerPoint or Canva presentation to present the project and results

Performance Criteria

  • ✔ Strict adherence to the steps
  • ✔ Quality of exploratory analysis
  • ✔ Statistical rigor
  • ✔ Pipeline with no data leakage
  • ✔ Correct model comparison
  • ✔ Interpretation of results
  • ✔ MAE objective < €15
  • ✔ Clean, structured, and reproducible code
  • ✔ Good team organization

Let me know if you need any adjustments or further assistance.

Edit

Pub: 21 Apr 2026 08:53 UTC

Views: 3