Data visualization with data refinery
Use data refinery on IBM Cloud Pak for Data to filter, clean and visualize data
This tutorial is part of the Getting started with IBM Cloud Pak for Data learning path.
|100||Introduction to IBM Cloud Pak for Data||Article|
|101||Virtualizing Db2 Warehouse data with data virtualization||Tutorial|
|201||Data visualization with data refinery||Tutorial|
|301||Data analysis, model building, and deploying with Watson Machine Learning with notebook||Pattern|
|401||Monitoring the model with Watson OpenScale||Pattern|
Data refinery is part of IBM Watson and comes with IBM Watson Studio on the IBM Public Cloud, and IBM Watson Knowledge Catalog running on premises using IBM Cloud Pak for Data. It’s a self-service data preparation client for data scientists, data engineers, and business analysts. With it, you can quickly transform large amounts of raw data into consumable, quality information that’s ready for analytics. Data refinery makes it easy to explore, prepare, and deliver data that people across your organization can trust.
After following this tutorial, the user will learn:
- How to Load data into the IBM Cloud Pak for Data platform for use with data refinery.
- How to transform a sample data set, either by entering R code in the command line or selecting operations from the menu.
- How to use Data Flow steps to keep track of your work.
- How to visualize the data with charts and graphs.
Completing this tutorial should take about 45 minutes.
Step 1: Load the billing.csv data into data refinery
Download the billing.csv file.
From the Project home, click on
+Add Data Set, and choose the billing.csv file.
Click on the newly added billing.csv file.
Data Refinery should launch and open the data like the image below:
Step 2: Refine your data
We’ll start out in the Data tab.
Transform your sample data set by entering R code in the command line or selecting operations from the menu. For example, type filter on the Command line and observe that autocomplete will give hints on the syntax and how to use the command:
Alternatively, hover over an operation or function name to see a description and detailed information for completing the command. When you’re ready, click Apply to apply the operation to your data set.
Highlight the TotalCharges column and click the
First, we notice that TotalCharges is a string. However, since it represents a decimal number, let’s convert the values to decimal. Choose the Operator
Convert Column Type:
Then pick Decimal for the type, and click
filter and choose the TotalCharges column from the drop down, then the Operator Is empty:
We can see there are only 3 rows with an empty value for TotalCharges:
It should be save to just drop these rows from the data set, so let’s do that.
Choose the Operation Remove empty rows for the TotalCharges column:
Finally, we can remove the CustomerID column, since that won’t be useful for training a machine learning model in a later Code Pattern. If you are doing this as part of the IBM Cloud Pak for Data Learning Path, choose the Remove operator, and then click
Step 3: Use Data Flow steps to keep track of your work
What if we do something we don’t want? We can undo (or redo) an action using the circular arrows:
As you refine your data, IBM Data Refinery keeps track of the steps in your data flow. You can modify them and even select a step to return to a particular moment in your data’s transformation.
To see the steps in the data flow that you have performed, click the Steps button. The operations that you have performed on the data will be shown:
You can modify these steps in real time and save for future use.
Step 4: Profile the data
Clicking on the Profile tab will bring up a quick view of several histograms about the data.
You can get insight into the data from the histograms:
Twice as many customers are month-to-month as either 2-year or 1-year contract.
More choose paperless billing, but around 40% still prefer a paper bill mailed out to them.
You can see the distribution of MonthlyCharges and TotalCharges.
From the Churn column, you can see that a significant number of customers will cancel their service.
Step 5: Visualize with charts and graphs
Choose the Visualizations tab to bring up an option to choose which columns to visualize. Click on the empty space for the Columns to Visualize where the image below says
Click here, choose TotalCharges, then Click on Visualize data when ready:
We first see the data in a histogram by default. You can choose other chart types. We’ll pick
Scatter plot next by clicking on it:
In the scatter plot, choose TotalCharges for the x-axis, MonthlyCharges for the y-axis, and Churn for the Color map:
Scroll down and give the scatter plot a title and sub-title if you wish. Click on the “gear” under
Actions to perform tasks such as Start over, Download chart details, Download chart image, or Global visualization preferences:
We see that we can do things in the Global visualization preferences for Titles, Tools, Color schemes, and Notifications. Let’s set the Color scheme to Vivid:
Now the colors for all of our charts will reflect this:
This tutorial showed you a small sampling of the power of IBM Data Refinery on IBM Cloud Pak for Data. The tutorial also explained how you can transform data using R code, at the command line, using various operations on the columns such as changing the data type, removing empty rows, or deleting the column altogether. Next, the tutorial explained that all the steps in our data flow are recorded, so you can remove steps, repeat them, or edit an individual step. The tutorial also showed how you can quickly profile the data to see histograms and statistics for each column. And finally, the tutorial explained how you can create more in-depth visualizations, and create a scatter plot mapping TotalCharges vs. MonthlyCharges, with the churn results highlighted in color.
This tutorial is part of the Getting started with IBM Cloud Pak for Data learning path. To continue the series and learn more about IBM Cloud Pak for Data, take a look at the next pattern, Data analysis, model building, and deploying with Watson Machine Learning with notebook.