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)

Fixing many-to-many relationships

We sometimes have to work with a data table that was created from a many-to-many merge. This is a merge where merge-by column values are duplicated on both the left and right sides. As we discussed in the previous chapter, many-to-many relationships in a data file often represent multiple one-to-many relationships where the one side has been removed. There is a one-to-many relationship between dataset A and dataset B, and also a one-to-many relationship between dataset A and dataset C. The problem we sometimes have is that we receive a data file with B and C merged, but with A excluded.

The best way to work with data structured in this way is to recreate the implied one-to-many relationships, if possible. We do this by first creating a dataset structured like A; that is, how A is likely structured given the many-to-many relationship we see between B and C. The key to being able to do this is in identifying a good merge-by column for the data on...