Session 1 — From Raw Data to Clean Data

Instructor: Stéphane Derrode, Centrale Lyon
Formation: Centrale Digital Lab @ Ecole Centrale Lyon
← Back to course index

---

📦 Download all session files — notebook, datasets, requirements

⬇ session1.zip

Contents: session1_titanic.ipynb · titanic.csv · port_info.csv · requirements.txt

---

Overview

Dataset	Titanic passenger records
Duration	3 hours
Format	Jupyter notebook + paper quiz (15 min)
Pandas required?	No — introduced in this session

In this first session, you will discover Pandas by practice on the Titanic dataset. Rather than a systematic lecture on the library, concepts are introduced as you need them to answer real questions about the data.

---

Learning objectives

By the end of this session, you will be able to:

• Load a CSV file and perform a first inspection of a DataFrame
• Detect and quantify missing values, and choose an appropriate imputation strategy
• Drop irrelevant columns and encode categorical variables
• Join two DataFrames with pd.merge()
• Aggregate data with groupby() and pivot_table()

---

Before the session — what you need to do

1. Set up your environment

Make sure Python 3.9+ is installed. Then install the required packages:

pip install pandas numpy matplotlib seaborn jupyter

2. Download the session files

Download the following files and place them in a folder called session1/: ⬇ session1.zip.

3. Launch Jupyter

jupyter notebook

Then open session1_titanic.ipynb.

4. Check your setup

Run the first code cell (the import cell). If no error appears, you are ready.

---

Session content

The notebook is divided into 5 blocks:

Block	Topic	Key functions
1	First contact with Pandas	read_csv, head, shape, dtypes, describe
2	Exploring missingness	isnull, info, bar chart of missing values
3	Cleaning & Encoding	fillna, drop, map, get_dummies
4	Merging DataFrames	pd.merge, join types
5	Aggregation & Grouping	groupby, agg, pivot_table, heatmap

Each block contains exercises (🏋️) with solutions hidden in collapsible cells.
Try to solve them before expanding the solution.

---

Quiz

A 15-minute paper quiz (closed book, no devices) will be held at the end of the session.
It covers the concepts introduced in the notebook:

• True/False questions on Pandas behaviour and join semantics
• Multiple choice on missing value strategies and encoding choices
• Short interpretation questions on groupby output and pivot tables

💡 Tip: Focus on understanding why each operation is done, not on memorising syntax. The quiz tests reasoning, not code recall.

---

Key concepts to remember

• Always work on a copy — never modify the raw DataFrame in place
• Understand missingness before imputing — the right strategy depends on the context
• Binary map vs one-hot encoding — use map for 2 categories, get_dummies for 3+
• Left join preserves all rows from the left — always verify shape after a merge

---

Coming up in Session 2

→ Session 2 — From Clean Data to Insight

Spotify Tracks dataset · Distributions · Correlations · PCA