ANALISIS SENTIMEN OPINI PUBLIK TERHADAP IMPLEMENTASI LAYANAN DIGITAL AKADEMIK MENGGUNAKAN ALGORITMA BIDIRECTIONAL LONG SHORT-TERM MEMORY DAN SUPPORT VECTOR MACHINE

Abstract

Evaluating student perceptions regarding digital service transformation in higher education provides essential insights for institutional governance enhancement. This study presents a comparative performance analysis of Deep Learning Bidirectional Long Short-Term Memory (BiLSTM) and Machine Learning Support Vector Machine (SVM) coupled with TF-IDF and Word2Vec feature extraction for classifying public sentiment (positive, neutral, negative) toward academic digital platforms. The corpus consists of 1,500 Indonesian text reviews subjected to standard NLP preprocessing pipelines including case folding, tokenization, slang word normalization, and stemming. Models are evaluated using accuracy, precision, recall, F1-score, and k-fold cross validation. Experimental findings establish that the BiLSTM architecture outperforms SVM, achieving 92.67% accuracy and 92.40% F1-score, whereas SVM attained 86.33% accuracy and 85.90% F1-score. BiLSTM dual-directional temporal dependency modeling demonstrates superior capability in handling contextual nuance, negation, and colloquial linguistic variations.