IMPLEMENTASI ALGORITMA SPECTRAL CLUSTERING UNTUK ANALISIS SENTIMEN

Abstract

Data mining is a study that collects, cleans, processes, analyzes and benefits from data. One of the techniques known in data mining is the Spectral Clustering technique. Spectral clustering is a technique that follows the Connectivity approach, where this method classifies points that are connected or directly adjacent. The purpose of this study is to determine the level of public sentiment towards the 2017 Jakarta Pilkada using the Spectral Clustering method. The test data was obtained from the scraping process on Twitter from October 1, 2016 to April 20, 2017. In this study, input data consisting of tweet data and output data were used in the form of sentiments that have been clustered into 3, namely positive, negative and neutral. Obtained 4571 negative data, 1899 neutral data and 1588 positive data. with the highest possible win rate in the first round on Ahok. In the second round with 2205 data, 604 positive tweets were obtained, 1123 neutral data, 479 negative data for negative tweets. In the second round Anies Baswedan received higher positive and lower negative responses than Candidate Ahok, so that the chances of winning against Anies Baswedan were higher than Ahok.