Learning Pandas 2, Second Edition. Master Data Wrangling, NLP, Geospatial Analysis, and Production ML Pipelines using pandas 2.3 (en Inglés) - Matthew Rosch

This book has been updated with Pandas 2.3, and it's exactly what ML engineers, data scientists and data engineers have been waiting for. It's a hands-on desk guide that's full of solutions, and it's the most up-to-date, production-ready book to the most widely used data manipulation library in the Python ecosystem.

This book covers all the big changes in Pandas 2.3, like Copy-on-Write semantics, PyArrow-backed types that save over 50% memory, the new default StringDtype, and the deprecated frequency aliases that are messing up time series pipelines everywhere. All the chapters are based on one growing application using a real Customer Churn dataset, so every technique is put into a context where you can trace it and use it in production.Once you've got the hang of pandas, you will be exploring deep into feature engineering with feature_engine and scikit-learn's set_output API, dealing with class imbalance with SMOTE and ADASYN, and doing distributed computing with Dask, as well as JIT-compiled custom functions with Numba and JAX. On top of that, you'll be able to handle full NLP pipelines from TF-IDF to LDA topic modelling, and geospatial analysis with GeoPandas.

It doesn't matter if you're building ML pipelines, scaling data infrastructure, or connecting pandas to TensorFlow, PyTorch, or JAX, this book will give you the practical depth and modern patterns to do it correctly on pandas 2.3 today, and stay forward-compatible with pandas 3.0 tomorrow.

Build memory-efficient pipelines using PyArrow backends and targeted dtype choices.

Write Copy-on-Write-safe assignment patterns that work on pandas 2.3 and 3.0.

Engineer rich ML features using ratios, bins, group statistics, and interaction terms.

Handle class imbalance with SMOTE, ADASYN, and quantified pandas-based profiling.

Scale datasets beyond RAM using Dask lazy evaluation and distributed cluster computing.

Accelerate custom scoring functions with Numba JIT and JAX-compiled batch operations.

Extract sentiment, topics, and clusters from raw text using TF-IDF and LDA pipelines.

Perform spatial joins, buffer analysis, and geocoding with GeoPandas and geopy.

Preserve named DataFrames throughout sklearn Pipelines using the set_output API.

Migrate confidently from legacy pandas patterns to pandas 2.3 production standards.

Getting Started with Pandas 2.3

Data Read, Storage, and File Formats

Indexing and Selecting Data

Data Manipulation and Transformation

Time Series and DateTime Operations

Performance Optimization and Scaling

Machine Learning with Pandas 2.3

Text Mining and NLP

Geospatial Data Analysis