The world of big data has been expanding at a lightning pace. With all this data available, knowing how to best use it can be a challenge. According to Data Ideology, a data and analytics consulting firm, 95% of businesses cite the need to manage unstructured data as a problem for their business.
Here’s how exploratory data analysis fits into the data science process and an example of how it works.
What is Exploratory Data Analysis?
Exploratory Data Analysis (EDA) is a specific approach and philosophy to data analysis that uses different techniques. Analysts investigate data using graphic representations and statistics before making assumptions.
The key objectives of exploratory data analysis include:
- Uncovering underlying structure and extracting variables.
- Exposing anomalies and testing underlying assumptions.
- Maximizing data set insights and determining optimal settings for factors.
- Recommending hypotheses.
- Overseeing the selections of the necessary techniques and tools.
- Finding additional streams for data collection.
While often used interchangeably with statistical graphics, they aren’t the same. Statistical graphics focus on specific techniques that are graphically based and hones in on one aspect of the data characterization.
In contrast, exploratory data analysis is an overall philosophy on dissecting and interpreting a data set, using several of the same techniques as statistical graphics.
There are four primary types of exploratory data analysis: univariate non-graphical, univariate graphical, multivariate non-graphical, and multivariate graphical.
Benefits of Exploratory Data Analysis
Exploratory data analysis enables you to prep data sets for deep analysis to provide context for business problems. Plus, it provides a firm set of features with statistical learning and sets the stage for more advanced data analysis.
Finding patterns using data can help validate business assumptions and enable sound decision-making. By using exploratory data analysis, an organization can have greater confidence in the data they’re using.
EDA also helps to refine the choice of feature variables that can also use machine learning.
Exploratory Data Analysis In Action
Now, let’s look at an example of how exploratory data analysis can be used.
Let’s say you’re an online clothing retailer. Your data set will include information like quantities, unit prices, description, stock numbers, customer IDs, invoice numbers and more.
You know that your customers buy a variety of clothing and footwear and that certain customers will buy more than one type of shoes or pants.
Using exploratory data analysis, you discover exactly how many people buy three pairs of shoes or five pairs of pants every year. However, your EDA may also tell you that there is a subset of customers who buy 20 pairs of shoes and 10 pairs of pants every year.
Now that you’ve discovered this outlying data, you can start examining the additional questions you’d like answered about these customers — who they are, what their buying patterns are, and so on.
With this new insight, you can:
- Further explore the existing data to look for patterns.
- Collect different data that can be explored at a later date.
- Create marketing campaigns to directly target these customers.
Pursuing a Career in Exploratory Data Analysis
For those looking to break into the field of exploratory data analysis, starting with an
With a strong foundation in place, you’ll be able to grow your skillset and move towards a more focused exploratory data analysis role.
Looking to get into the field of exploratory data analysis?
The Udacity Data Analyst Nanodegree Program teaches learners how to use Python, SQL and statistics to uncover insights, communicate critical findings, and create data-driven solutions.