Support Vector Classification with Hyperparameters for Analysis of Public Sentiment on Data Security in Indonesia

Abstract

The development of Information Technology makes increasing use of the internet. This raises the vulnerability of data security. Cyber attacks in Indonesia caused many tweets on social media Twitter. Some are positive, and some are negative. The problem of this study is to determine the public sentiment towards data security in Indonesia, while the purpose of this study is how the response or evaluation of the government of Indonesia to the many perceptions of people who lack confidence in data security in Indonesia. Data obtained from twitter with as much as 706 data was processed using python with a percentage of 10% test data and 90% training data. Weighting is done using TF-IDF, and then the Data is processed using the Support Vector Machine algorithm using the SVC (Support Vector Classification) library. Support Vector Classification with RBF kernel classifies Text well to obtain AUC value with good classification category. Utilizing one of the hyperparameter techniques, which is a grid search technique that can compare the accuracy of test results. The test results using SVC with RBF kernel obtained an accuracy value of 0.87, Precision of 0.82, recall of 0.94, and F1_Score of 0.87. This study is expected to be used by decision-makers related to public confidence in data security in Indonesia