This project explores the use of Extreme Gradient Boosting (XGB) Regression to predict employee promotions, focusing on continuous variable outcomes rather than binary classifications. While classification models predict whether an employee will or will not be promoted, XGB Regression provides a nuanced score that reflects the likelihood of promotion, allowing for more detailed insights into employee performance and potential.
Project Overview
XGB Regression is designed to handle continuous variables, making it ideal for predicting outcomes that vary across a spectrum rather than fitting into simple "yes" or "no" categories. In the context of promotion prediction, the model generates a numerical score indicating how likely an employee is to be promoted based on various features like performance scores, qualifications, and training history.
While the model achieved a lower performance accuracy of 34%, this outcome reflects the complexity of interpreting continuous scores compared to classification outputs. The model’s focus on delivering nuanced, continuous outcomes presents both challenges and opportunities for improvement in HR decision-making.
Key Features
Continuous Variable Prediction:
XGB Regression predicts a continuous score for each employee, offering a range of potential promotion outcomes rather than a binary yes/no result.
Nature of Predictions: The model outputs scores such as 40, 75, or 90, reflecting varying degrees of promotion likelihood. These scores require careful interpretation to guide promotion decisions.
Data Overview:
Training Data: Employee data including 19 features such as qualifications, job performance, and training scores were used to train the model.
Testing Data: A separate test set with 18 features was used to evaluate the model’s predictive performance, excluding the promotion outcome for unbiased validation.
Current Performance:
The model achieved 34% accuracy, which is lower than expected for promotion prediction. This is due to the complexity of continuous variable prediction, where the output is not a clear-cut binary result.
This performance highlights the challenge of using regression in contexts where classification might provide more direct answers.
Challenges with Continuous Outcomes:
Complexity in Interpretation: Unlike classification models, which provide a straightforward yes/no prediction, XGB Regression outputs require more nuanced interpretation. For example:
A score of 75 might suggest a strong likelihood of promotion.
A score of 40 could indicate that the employee needs further development.
Stakeholder Impact: HR professionals must interpret these scores and decide how to act on them, making the model’s results less definitive and more open to subjective analysis.
Opportunities for Improvement:
Refinement of the Model: The relatively low accuracy presents a chance to improve the model through techniques like:
Feature Engineering: Creating new features or refining existing ones to provide more predictive power.
Hyperparameter Tuning: Optimizing the model’s parameters to improve performance.
Continuous Training: Incorporating new data over time to keep the model updated and relevant.
Challenges and Solutions
Interpretation of Continuous Variables: One of the main challenges is translating the continuous output of the regression model into actionable decisions. While a score provides a nuanced view, HR teams must determine the threshold at which an employee is deemed promotable.
Lower Accuracy: The model’s 34% accuracy reflects the difficulty of applying regression to a task better suited to classification. This challenge can be addressed through further refinement and optimization of the model’s features and parameters.
Final Deliverables
XGB Regressor Model: A model trained to predict the likelihood of employee promotion using continuous variables, offering nuanced predictions that go beyond binary outcomes.
Performance Reports: Accuracy reports highlighting the model’s current performance level and areas for future improvement.
Feature Importance Analysis: Insights into which features (e.g., qualifications, performance scores) had the most influence on the model’s predictions.
Conclusion
The XGB Regression model provides a more detailed view of employee promotion likelihood by predicting continuous outcomes, offering deeper insights into performance metrics. While this approach adds complexity in terms of interpretation, it also opens the door for more refined and informed HR decision-making. As the model evolves through continuous training and optimization, it has the potential to become a powerful tool in promotion prediction.
GitHub Link: View Full Code
