Discovery : Jurnal Ilmu Pengetahuan
Vol 11 No 2 (2026): October 2026

Analisis Sentimen Program Magang Generasi Bertalenta Badan Usaha Milik Negara (MAGENTA BUMN) di Media Sosial X Menggunakan Word2Vec dan Support Vector Machine (SVM)

Aulivia Gustiniar Fitrianavrisha (Unknown)
Nilogiri, Agung (Unknown)
Umilasari, Reni (Unknown)



Article Info

Publish Date
22 Oct 2026

Abstract

The State-Owned Enterprises (SOEs) Talent Generation Internship Program (Magenta BUMN) is a human resource development program managed by the Indonesian Human Capital Forum (FHCI) that has attracted considerable public attention on the social media platform X. The high volume of unstructured public opinions makes manual program evaluation challenging. This study aims to measure the model's accuracy, precision, recall, and F1-score; evaluate the effectiveness of combining the Word2Vec Skip-gram architecture with a Linear-kernel Support Vector Machine (SVM) to classify public sentiment toward Magenta BUMN; and identify the dominant aspects of positive and negative sentiment. The dataset consists of 664 Indonesian-language tweets collected between January 2023 and November 2025. The data underwent six preprocessing stages and were evaluated using 18 parameter combinations involving variations in data-splitting ratios, vector size, and window size. The best performance was achieved with an 80:20 data-splitting ratio, a vector size of 300, and a window size of 7, yielding an accuracy of 80.45%, precision of 78.46%, recall of 80.95%, and an F1-score of 79.69%. The average cosine similarity of 0.8991 indicates that the model effectively captures contextual relationships between words within the Magenta BUMN domain. Positive sentiment is predominantly associated with career development and financial compensation, whereas negative sentiment primarily concerns selection transparency and technical issues with the platform.

Copyrights © 2026