Introduction
Hi, I am Akira, the editor-in-chief of Data Without Code. In our previous Use Case, we explored the fascinating world of unstructured data by performing social media sentiment analysis. You can now automatically read the minds of your customers.
But what about the minds of your own employees? As a DX manager, I have seen companies spend millions on marketing, only to lose their best talent because HR didn’t realize someone was unhappy until they handed in their resignation letter. By the time an exit interview happens, it is already too late.
What if you could predict who is going to quit before they actually do? In the data science world, this is called Employee Churn or Turnover Prediction. And just like our sales forecasting model, you do not need to be a Python programmer to build it.
In this tutorial, I will show you how to apply HR analytics in KNIME to build an employee turnover prediction model using a simple, visual Decision Tree.
Why Use a Decision Tree for HR?
When dealing with human resources, you cannot just tell a manager: “The computer says Akira is going to quit.” The manager will immediately ask: “Why?”
Some machine learning models are “black boxes,” meaning they give you an answer but don’t explain how they got it. A Decision Tree is different. It creates a visual flowchart that humans can easily read. It might say: “IF the employee has been here for 3 years, AND their salary has not increased, AND they work over 50 hours a week, THEN they have an 85% chance of quitting.”
This transparency makes Decision Trees the perfect algorithm for HR analytics.
Step 1: Gather Your HR Data
To train our model, we need historical data of employees who stayed and employees who left. You can pull this directly from your HR system using the SQL Database Connector or simply read an exported CSV using the CSV Reader node.
Your dataset should include columns like:
- Years at company
- Current salary
- Overtime hours
- Time since last promotion
- Left Company? (Yes/No) – This is our “Target Column” that we want the model to learn.
Step 2: Train the Model (Decision Tree Learner)
Now, we let KNIME find the hidden patterns in your historical HR data.
- Search for the Decision Tree Learner node in your Node Repository and drag it to your canvas.
- Connect your historical HR data to the node.
- Double-click to configure it. At the very top, select your Target Column: “Left Company?”.
- Execute the node.
That’s it! If you right-click the executed node and select “View: Decision Tree View,” a new window will open displaying a beautiful, interactive flowchart. You can literally trace the path of why past employees decided to leave.
Step 3: Predict the Future (Decision Tree Predictor)
The model has learned the rules. Now, let’s apply those rules to your current active employees to see who is at risk right now.
Grab your dataset of current employees (who obviously have “No” in the “Left Company?” column right now).
- Add a Decision Tree Predictor node to your canvas.
- Connect the blue square output (the trained model) from the Learner node to the blue square input of the Predictor node.
- Connect your current employee data to the standard black triangle input of the Predictor node.
- Double-click the Predictor. Check the box that says “Append class probabilities”. This tells KNIME to output a percentage risk, not just a simple Yes/No.
Execute the node. When you view the output, you will see a brand new column showing the exact probability of each active employee quitting. If someone shows a 90% risk, HR can proactively schedule a 1-on-1 meeting or offer a promotion before it is too late!
Conclusion: Your Next Steps
Congratulations! You have just built an enterprise-grade predictive HR model. By combining historical data with the Decision Tree nodes, you have transformed your HR department from reactive to proactive.
If you want to make this even more powerful, you can use the Top k Selector node to extract the Top 10 highest-risk employees and automatically email that list to the HR Director every Monday morning.
We have covered sales, supply chain, and HR. Now, let’s look at digital marketing and website traffic. If you use Google Analytics but struggle to tie that web traffic back to your actual CRM sales, I have the perfect solution for you.
Join me in our next Use Case tutorial: Web analytics: Blending Google Analytics data with CRM data in KNIME!
