Optimasi Analisis Sentimen Pada Twitter Olshop Tokopedia Menggunakan Textmining Dengan Algoritma Naïve Bayes & Adaboost

Abstract

Sentiment Analysis or commonly called Opinion Mining is the process of understanding, extracting and processing textual data automatically to obtain sentiment information contained in a sentence of opinion or opinion on a problem or object by someone, whether it tends to have a positive or negative opinion. This study aims to classify tweet data into 2 classifications, namely positive and negative. In this study, Indonesian text is used on Twitter social media in the form of tweets related to Tokopedia. Public opinion contained in the tweet can be used as material to find out whether tweets on Twitter, especially on Tokopedia, are classified as positive or negative. The data used consists of 1,000 tweet data. This dataset comes from the tweets of Tokopedia customers written on the Tokopedia twitter account. In text mining techniques, “transform case”, “tokenize”, “token filter by length”, “stemming” are used to build classifications. Gataframework is used to help during the preprocessing and cleansing process. RapidMiner is used to help create sentiment analysis in comparing three different classification methods, on Tokopedia's tweet data. The method used to compare in this research is the Naïve Bayes algorithm and the Naïve Bayes algorithm which is added with the Synthetic Minority Over-sampling Technique (SMOTE) feature and the Naïve Bayes algorithm is added with the Synthetic Minority Over-sampling Technique (SMOTE) feature which is optimized with Adboost. . The Naïve Bayes algorithm added with the Synthetic Minority Over-sampling Technique (SMOTE) feature, which was optimized with Adboost, got the best score. With 94.95% accuracy, 90.86% precision, 100.00% recall and 0.950 AUC