Understanding Employees’ Capability Towards Workload Clustering From E-Portfolio Data

This paper proposed the workload clustering method for extracting a frequent task assigned to the employees. The data was exported from an Electronic Portfolio (e-Portfolio) system. In this work an investigation on the effectiveness of workload clustering through various clustering algorithms and dimension-reduction techniques were conducted. This research deployed Natural Language Processing for word tokenizing and feature engineering. The Silhouette scores was adopted to measure across different clustering algorithms such as K-means, Hierarchical, Gaussian Mixture Model (GMM), Fuzzy C-means (FCM), and DBSCAN. As a result, the hierarchical clustering on three clusters achieved the best Silhouette score of 0.442. Subsequently, after performing the Principal Component Analysis (PCA) and visualizing the clustering results in 2-dimensional space, the criteria on clear separation, and distinct employees’ capability were observed.