Oppiskele Evaluating Model Performance | Predictive People Analytics

Pyyhkäise näyttääksesi valikon

Predictive models play a crucial role in People Analytics by helping you anticipate important outcomes such as employee attrition, promotion, or engagement. To trust and effectively use these models, you need to understand how well they perform. The most common metrics for evaluating classification models in HR contexts are accuracy, precision, and recall.

Accuracy measures the proportion of correct predictions out of all predictions made. In HR, this might tell you how often your attrition model correctly predicts whether an employee will stay or leave. Precision focuses on the correctness of positive predictions—how many employees predicted to leave actually do leave. This is vital if the cost of a false alarm (predicting someone will leave when they do not) is high. Recall measures how many actual cases of attrition your model successfully identifies. High recall is important when missing a true case (such as not identifying an at-risk employee) could have serious consequences.

Selecting the right metric depends on your business priorities. For example, in attrition prediction, you might prioritize recall to ensure you identify as many potential leavers as possible, even if it means a few false positives.


              12345678910111213
            
from sklearn.metrics import accuracy_score, precision_score, recall_score

# Simulated predictions from an attrition model
y_true = [0, 1, 0, 1, 0, 1, 1, 0, 1, 0]  # 1 = left, 0 = stayed
y_pred = [0, 1, 0, 0, 0, 1, 1, 0, 1, 1]  # model predictions

accuracy = accuracy_score(y_true, y_pred)
precision = precision_score(y_true, y_pred)
recall = recall_score(y_true, y_pred)

print("Accuracy:", round(accuracy, 2))
print("Precision:", round(precision, 2))
print("Recall:", round(recall, 2))

To gain deeper insights into your model's predictions, you can use a confusion matrix. This tool breaks down the results into four categories: true positives, true negatives, false positives, and false negatives. For HR decisions, the confusion matrix helps you see not only how many predictions were correct, but also what types of mistakes your model is making. For instance, if your attrition model has many false negatives, you are missing employees who are actually at risk of leaving—which could lead to costly surprises. On the other hand, too many false positives might mean unnecessary interventions for employees who were never planning to leave.

Interpreting the confusion matrix allows you to adjust your strategy, such as changing the threshold for flagging at-risk employees or focusing on improving recall or precision, depending on your organization's goals.


              123456789101112
            
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay
import matplotlib.pyplot as plt

# Simulated predictions
y_true = [0, 1, 0, 1, 0, 1, 1, 0, 1, 0]
y_pred = [0, 1, 0, 0, 0, 1, 1, 0, 1, 1]

cm = confusion_matrix(y_true, y_pred)
disp = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=["Stayed", "Left"])
disp.plot(cmap=plt.cm.Blues)
plt.title("Attrition Model Confusion Matrix")
plt.show()