The Drivers tab is super useful to draw meaningful insights from your prediction. It also lists the most important features/columns that were used to predict the output variable/target column.
The Drivers tab has two elements
- Drivers for predicting the target
- Top Drivers by impact
The following figures show the Drivers tab.
Drivers for Predicting the Target
The grey circles represent the features/columns and their position reflecting their relationship with the prediction column. A grey circle on the right shows a positive relationship between that feature and the prediction column. On the other hand, the grey circle on the left represents a negative correlation with the predictions column.
Let us consider a case study that involves predicting Customer Churn.
You can move the cursor on the circle to find out which column that particular circle represents.
The blue circle represents the column 'Total charges'. This means that 'Total Charges' is directly proportional and has a positive correlation with Churn.
Top Drivers by Impact
The "Top Drivers by Impact" give a visual representation of how a particular column impacts the target variable.
Let us understand how InternetService impacts customer Churn.
- Navigate to the target drop-down menu to select the target.
- Click on the drop-down button and select Churn = Yes, which refers to the customers who have churned and have cancelled their service.
- Finally, click on the InternetService column.
- This will give a Bar Chart representing the different categories and the impact they have on Churn.
- Move the cursor on the individual bars to find the exact value of the predicted churn percentage.
- This gives some great insights on how InternetService impacts Churn. We can clearly conclude from the plot that the customers who opted for 'Fiber Optics' have a high predicted Churn percentage (~42%). On the other hand, the customers who opted for DSL have a lower predicted Churn percentage (~20%).