<![CDATA[Alumni career mapping is a crucial aspect of evaluating and developing higher education programs. Cluster analysis, particularly the integration of k-means and autoencoder methods, has emerged as an effective solution for grouping complex and multi-dimensional alumni career data. This study aims to implement and assess the combination of k-means and autoencoder algorithms in alumni career mapping based on GPA, study duration, waiting time, job type, salary, job level, and field of study suitability. The autoencoder is employed to reduce dimensions, while k-means clusters alumni into groups based on the similarity of their career profiles. The data used in the cluster analysis is sourced from the tracer study. Pre-processing of the tracer study data is conducted through several stages, including cleaning, encoding, and normalization. The evaluation results indicate that the combination of k-means and autoencoder yields superior Silhouette and DBI scores. The Silhouette score with the autoencoder achieved 0.6112, while without it, the score was only 0.3956. The DBI value with the autoencoder is 0.566, whereas without it, the DBI reached 1.022. This cluster analysis effectively grouped the tracer study data into six clusters based on similarities in career profiles. The clustering results suggest that the formed clusters are more influenced by the alumni's job type and duration of study.]]>
Copyrights © 2025
