Optimizing Learning Pathway in Human Capital Development Programs with BERT Transformer

Abstract

This study presents a novel approach to assist learning analysts in identifying suitable learning pathways based on historical training data through the utilization of text mining techniques. The dataset utilized in this research comprises training data from the year 2021 and the Course Development Management Program (CDMP) catalogue. The BERT 'bert-base-nli-mean-tokens' model is employed for encoding purposes. By comparing the training data names from 2021 with the CDMP catalogue using cosine similarity and dot score, valuable insights are obtained. The findings indicate that cosine similarity is a more effective measure for interpreting the data, thereby simplifying the process for learning analysts and managers in identifying appropriate learning paths for their employees. This research provides a practical solution that leverages text mining techniques to optimize the analysis and decision-making processes in learning and development domains, enabling organizations to enhance the effectiveness and efficiency of their training programs.