Python Data Science Cookbook

This is the Chinese translation of Python Data Science Cookbook, by Gopi Subramanian, a data scientist with more than a decade of experience in data mining and machine learning. The book follows a recipe format: each task first sets out the "getting ready" section, then the "how to do it" and "how it works" sections, and finally extends with "there's more" and "see also," so readers can dip in as needed or work through it systematically from the start. It begins with Python fundamentals, covering data structures and programming techniques commonly used in data science—dictionaries, dictionaries of dictionaries, tuples, sets, list comprehensions, iterators, and generators—then introduces how to set up a Python scientific computing environment. Two chapters on data analysis follow, covering data exploration, descriptive statistics, data visualization, and deeper data processing. The book then moves into data mining topics, discussing feature processing and similarity and distance measures—problems of the "needle in a haystack" kind. The second half turns to machine learning: linear regression, logistic regression, regularization and variable selection, classification and clustering, and model evaluation methods. It then covers ensemble methods (bagging, boosting, voting), random forests, rotation forests, extremely randomized trees, and other tree-growing models, closing with "large-scale machine learning—online learning," which introduces the perceptron, stochastic gradient descent for regression and classification, and how to model data that cannot fit into memory all at once in a streaming fashion. The examples span numeric and text datasets and balance mathematical principles with runnable code, making the book a useful desk reference for data science algorithms. Because the original was published in 2015, the code is based on the numpy, scipy, scikit-learn, and matplotlib ecosystem of that time, and readers should watch for version differences.