Analisis Sentimen Masyarakat Mengenai Revisi Undang-Undang Desa Pada Media Sosial Twitter Dengan Menggunakan Metode Naive Bayes Dan Feature Selection Particle Swarm Optimization
P N Harun, Muhtajudin Danny, Zy A T · Zenodo (CERN European Organization for Nuclear Research) · 2023
The ratification of Law no. 6 of 2014 concerning Villages (hereinafter referred to as the Village Law) on 15 January 2014 is a historic milestone in the history of village policies. According to the dpr.go.id page, thousands of Village Heads held a demonstration on January 17 2023 demanding revisions to Law Number 6 of 2014 concerning Villages, in the courtyard of the DPR RI Building. One of the aspirations conveyed by the protesters was regarding the term of office of the head village which was extended to nine years. This reason became a factor in the pro and contra comments regarding this village law which reaped many responses in the form of tweets from various groups of people, resulting in many traces of tweets which contained public opinion regarding the revision of the 2023 village law on Twitter social media. This study aims to determine the results of the classification of public sentiment regarding the Village Law on Twitter social media and to determine the results of accuracy, precision, recall resulting from the use of the Naïve Bayes method and the Particle Swarm Optimization feature in RapidMiner Studio software. Naïve Bayes Classifier is a machine learning method that uses probability calculations. Particle Swarm Optimization is an optimization method inspired by the behavior of fish and poultry schools in searching for food sources. The preprocessing stage in this study includes cleansing, removing duplicates, data selection, normalization, case folding, tokenizing, filtering, stopwords, stemming, and labeling. The classification results obtained were 52.15% of Twitter users commented positively and 47.85% of Twitter users commented negatively about the Village Law. The accuracy value obtained increased by 4.18% from 72.53% to 76.71%, the precision value obtained increased by 4.13% from 72.22% to 76.35%, and the recall value obtained increased by 4 .98% from 72.50% to 77.48%.