Book Image

Python Data Cleaning Cookbook

By : Michael Walker
Book Image

Python Data Cleaning Cookbook

By: Michael Walker

Overview of this book

Getting clean data to reveal insights is essential, as directly jumping into data analysis without proper data cleaning may lead to incorrect results. This book shows you tools and techniques that you can apply to clean and handle data with Python. You'll begin by getting familiar with the shape of data by using practices that can be deployed routinely with most data sources. Then, the book teaches you how to manipulate data to get it into a useful form. You'll also learn how to filter and summarize data to gain insights and better understand what makes sense and what does not, along with discovering how to operate on data to address the issues you've identified. Moving on, you'll perform key tasks, such as handling missing values, validating errors, removing duplicate data, monitoring high volumes of data, and handling outliers and invalid dates. Next, you'll cover recipes on using supervised learning and Naive Bayes analysis to identify unexpected values and classification errors, and generate visualizations for exploratory data analysis (EDA) to visualize unexpected values. Finally, you'll build functions and classes that you can reuse without modification when you have new data. By the end of this Python book, you'll be equipped with all the key skills that you need to clean data and diagnose problems within it.
Table of Contents (12 chapters)

Classes that contain the logic for updating series values

We sometimes work with a particular dataset for an extended period of time, occasionally years. The data might be updated regularly, for a new month or year, or with additional individuals, but the data structure might be fairly stable. If that dataset also has a large number of columns, we might be able to improve the reliability and readability of our code by implementing classes.

When we create classes, we define the attributes and methods of objects. When I use classes for my data cleaning work, I tend to conceptualize a class as representing my unit of analysis. So, if my unit of analysis is a student, then I have a student class. Each instance of a student created by that class might have birth date and gender attributes and a course registration method. I might also create a subclass for alumni that inherits methods and attributes from the student class.

Data cleaning for the NLS DataFrame could be implemented nicely...