To make a prediction for an employee, the prediction algorithm:
- Looks at the employee's attributes and determines the employee’s path for each tree.
- Averages the historical event likelihoods across the trees.
Example: Predicting the likelihood that Floyd McGregor will resign in the next 12 months.
The following table lists some of Floyd’s attributes.
For this example, our predictive model is composed of three decision trees. To predict Floyd’s overall risk of resignation, the prediction algorithm looks at Floyd’s attributes and determines the path he falls into for each tree. The event likelihood for each tree is based on the chosen path and the population of employees who resigned at the leaf node. An overall risk of resignation is calculated for Floyd by averaging the historical event likelihoods across the trees. The following illustration shows how the Predicted Risk of Resignation model predicts a 35 percent likelihood that Floyd will resign in the next 12 months.