Informative Tweet Classification of the Earthquake Disaster Situation In Indonesia

Twitter is one of the online media that produces a number of information data on significant disaster situations, namely tweets, some of which have tweets that have valuable information to take humanitarian assistance measures. The purpose of this study is to train models based on information sourced from Twitter by classifying tweets that are informative and not informative by using algorithms for classification. Based on previous research algorithms that are considered appropriate in dealing with this problem are Support Vector Machine algorithms based on that, in research developing research models by adding features of the Smote UpSampling imbalance and Gini Index and adding a Naïve Bayes algorithm to compare the accuracy of the classification algorithm. The data used is the data of tweets related to earthquake disaster events that occurred in Indonesia that were collected using the Rapid Miner application and using GataFramework in text preprocessing. Based on the method proposed in this study the Support Vector Machine algorithm produces 81.03% accuracy superior to the Naïve Bayes algorithm which produces an accuracy of 80.30%, based on the results of accuracy both enter into a good classification algorithm of the resulting accuracy.