Data Science for Economics (Practicals)

Schedule and location

• Session #1: Thursday, October 2, 9:30 AM - 11:00 AM @ Room M1 (CEMFI)
• Session #2: Tuesday, October 27, 15:00 PM - 16:30 PM @ Room M1 (CEMFI)
• Session #3: Thursday, November 19, 9:30 AM - 11:00 AM @ Room M1 (CEMFI)

Please bring your laptop to class with VSCode and uv installed and ready to run Python code.

Repository structure

.
├── data/                          # Data files for exercises and examples
│   ├── all_ECB_speeches.csv
│   ├── death-rate-who.csv
│   ├── gdp-per-capita-worldbank.csv
│   └── population-unwpp.csv
├── Session1/                      # Session 1 notebooks
│   ├── 1_1.ipynb                 # Introduction to Python (Part I)
│   └── 1_2.ipynb                 # Introduction to Python (Part II)
├── Session2/                      # Session 2 materials
│   ├── 2_1_Scraping_Intro.ipynb  # Web scraping introduction
│   ├── 2_2_BIS_Scraper_I.ipynb   # BIS scraper (Part I)
│   ├── 2_3_BIS_Scraper_II.ipynb  # BIS scraper (Part II)
│   ├── 2_4_Forecasting_Professions.ipynb  # Forecasting professions exercise
│   ├── Extra_BCU_Scraper.ipynb   # BCU scraper (extra)
├── Session3/                      # Session 3 materials
│   ├── 3_1_LLM_Intro.ipynb       # Introduction to Large Language Models
│   ├── 3_2_LLM_Fine_Tuning.ipynb  # Fine-tuning LLMs
│   ├── 3_3_LLM_Function_Calling.ipynb  # LLM function calling
├── Installation_Guide_VSCode_uv.pdf  # Installation guide
├── main.py                        # Main Python script
├── pyproject.toml                # Project dependencies (uv)
├── uv.lock                       # Locked dependencies for reproducibility
└── README.md                     # This file

Session plan

Session 1: Introduction to Python

1_1.ipynb - Introduction to Python | Part I

• Overview of Python and setup with UV package manager
• Jupyter Notebooks fundamentals
• Python syntax basics:
  • Variables and naming conventions
  • Primitive data types (int, float, string, bool, None)
  • Container data types (tuples, lists, dictionaries)
  • Functions and control flow (if/elif/else, for/while loops)
• Introduction to NumPy:
  • Creating and managing arrays
  • Array operations (element-wise and matrix multiplication)
  • Indexing and slicing

1_2.ipynb - Introduction to Python | Part II

• Introduction to Pandas:
  • Series: creation, operations, and slicing
  • DataFrames: creation, indexing, and operations
  • Data manipulation (adding/removing columns, sorting, merging)
  • Applying functions
• Data visualization:
  • Matplotlib basics
  • Seaborn for statistical graphics
  • Creating plots (scatter, line, bar, histograms, heatmaps)

Session 2: Web Scraping

2_1_Scraping_Intro.ipynb - Scrape it like you mean it!

• Introduction to web scraping:
  • Definition and use cases
  • Ethical considerations and legal aspects
  • HTTP status codes and requests
• Tools and libraries:
  • Using requests for HTTP requests
  • Parsing HTML with BeautifulSoup
  • Basic scraping patterns

2_2_BIS_Scraper_I.ipynb - Scraping Central Bank Speeches from the BIS (Part I)

• Understanding website structure and dynamic content
• Handling pagination in web scraping
• Building a scraper for BIS central bank speeches:
  • Extracting speech metadata
  • Downloading PDF documents
  • Organizing scraped data

2_3_BIS_Scraper_II.ipynb - Scraping Central Bank Speeches from the BIS (Part II)

• Advanced scraping techniques:
  • Using Selenium for dynamic web pages
  • PDF text extraction with PyPDF2
  • Handling file downloads and organization
• Setting up project structure:
  • Directory management
  • Data organization and storage

2_4_Forecasting_Professions.ipynb - Who is the Econometrician in the room?

• Combining web scraping with machine learning:
  • Image retrieval and processing
  • Profession classification using ML models
  • Forecasting applications

Extra_BCU_Scraper.ipynb - BCU Scraper (Extra)

• Additional practice with web scraping
• Scraping from Banco Central del Uruguay (BCU) website

Session 3: Large Language Models (LLMs)

3_1_LLM_Intro.ipynb - Introduction to Large Language Models

• Fundamentals of LLMs:
  • What are Large Language Models?
  • Transformer architecture overview
  • Pre-trained models and their applications
• Working with LLMs:
  • Using the transformers library
  • Text generation and completion
  • Model visualization with bertviz

3_2_LLM_Finetuning.ipynb - Fine-Tuning Large Language Models

• Fine-tuning techniques:
  • Parameter-Efficient Fine-Tuning (PEFT)
  • LoRA (Low-Rank Adaptation) method
  • Training custom models on domain-specific data
• Practical implementation:
  • Preparing datasets for fine-tuning
  • Training configuration and hyperparameters
  • Model evaluation and checkpointing
• Applications:
  • Text classification tasks
  • Domain adaptation for economic/financial texts

3_3_LLM_II_Function_Calling.ipynb - Structured Data with LLMs Done Right

• Function calling and structured output:
  • Defining tool schemas for LLM function calls
  • Implementing function calling patterns
  • Validating and executing LLM-requested functions
• Practical applications:
  • News → firm-level shocks: extracting affected firms and classifying shock types
  • Central bank speeches → policy stance classification and economic indicator extraction
  • Structured data extraction from unstructured text

Prerequisites

• Python 3.11+
• VSCode
• UV package manager installed with Python

Getting started

1. Clone this repository
2. Navigate to the project directory
3. Install dependencies: uv sync
4. Open notebooks in VSCode and select the UV environment

Contact

• Email: jesus.villota@cemfi.edu.es
• LinkedIn: https://www.linkedin.com/in/jesusvillotamiranda/