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)

Using stack and melt to reshape data from wide to long format

One type of untidiness that Wickham identified is variable values embedded in column names. Although this rarely happens with enterprise or relational data, it is fairly common with analytical or survey data. Variable names might have suffixes that indicate a time period, such as a month or year. Another case is that similar variables on a survey might have similar names, such as familymember1age, familymember2age, and so on, because that is convenient and consistent with the survey designers' understanding of the variable.

One reason why this messiness happens relatively frequently with survey data is that there can be multiple units of analysis on one survey instrument. An example is the United States decennial census, which asks both household and person questions. Survey data is also sometimes made up of repeated measures or panel data, but nonetheless often has only one row per respondent. When this is the case...