Design of an Integrated Model Using Machine Intelligence with AutoML for Employee Performance Classification

Abstract

As the dependence of organizations on data-driven decision-making for talent management is ever increasing, there is an increasing demand for developing accurate and interpretable employee performance classification models. Traditional methods will normally fail to deliver enough accuracy when a model's interpretability is guaranteed or vice versa; therefore, they may require highly manual tuning that would hamper scalability and generalizability. Therefore, to overcome the drawbacks of these methods, the paper herein proposes a hybrid scheme based on techniques offered by Random Forest, Gradient Boosting Machine, and AutoML in classifying and ranking employee performance effectively. In their selection, Random Forest is used because it can address complex structured data while producing clear feature importance insights. It incorporated GBMs to improve further accuracy in classification by iteratively correct model errors and resulted in more refined rankings. Finally, the usage of AutoKeras and H2O AutoML is to automatically test various algorithms and hyperparameters to select the best performing model that can emerge. This results in both high interpretability as well as superior predictive accuracy. In addition, layered analysis of the fused model is used for improvement in several ways at performance classification. Random Forests provide an initial kind of classification that comes with feature importance analysis, which helps the manager understand what drives key employee performance. GBMs improve ranking accuracy by removing residual errors, whereas AutoML optimizes the model configuration by using automated hyperparameter tuning and model selection. In this work, the achieved accuracy was 94-95%, which was above the threshold of the manually tuned models and provided actionable insights into the performance of the employees. The proposed model not only improves the classification but also improves the interpretability of the model, allowing organizations to make strategic decisions in talent management by leveraging data for recommendation. This integration of approaches fills this gap between model accuracy and pragmatic usability.