PENERAPAN LEXICON-BASED LABELING UNTUK ANALISIS SENTIMEN ULASAN UMKM KULINER DENGAN PERBANDINGAN KINERJA ALGORITMA MACHINE LEARNING

Abstract

This study aims to classify consumer sentiment of culinary MSMEs in South Kalimantan using a hybrid approach that combines Lexicon-Based methods and Machine Learning algorithms. The dataset used is a collection of MSME consumer reviews written on social media Instagram and Twitter (X). Lexicon-based is used for automatic labeling based on positive and negative word dictionaries, while Machine Learning algorithms (Naïve Bayes, Random Forest, Logistic Regression, Support Vector Machine (SVM), Decision Tree, and K-Nearest Neighbors (KNN)) as a model to determine the results of pattern-based sentiment classification. The research stages include dataset collection, text preprocessing, automatic labeling using lexicon-based labeling, TF-IDF feature extraction, model implementation with Machine Learning and evaluation of results using classification metrics (accuracy, precision, recall and F1-score), then comparative analysis between models to determine which model has the best performance in classifying MSME consumer review sentiment. The results showed that the Support Vector Machine model produced the best accuracy (79%) and an F1-score of 70%, followed by Logistic Regression with 77% accuracy and 77% precision. Both models performed better in classification than Naive Bayes and Random Forest, while the tree-based model was considered suboptimal. The results of this study are expected to contribute and serve as a reference for culinary MSMEs in East Kalimantan in understanding consumer perceptions to improve product and service quality