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)

Developing a merge routine

I find it helpful to think of merging data as the parking lot of the data cleaning process. Merging data and parking may seem routine, but they are where a disproportionate number of accidents occur. One approach to getting in and out of parking lots without an incident occurring is to use a similar strategy each time you go to a particular lot. It could be that you always go to a relatively low traffic area and you get to that area the same way most of the time.

I think a similar approach can be applied to getting in and out of merges with our data relatively unscathed. If we choose a general approach that works for us 80 to 90 percent of the time, we can focus on what is most important – the data, rather than the techniques for manipulating that data.

In this recipe, I will demonstrate the general approach that works for me, but the particular techniques I will use are not very important. I think it is just helpful to have an approach that...