Sentiment Analysis of Pedulilindungi Application Reviews Using Naive Bayes Classifier and Support Vector Machine

The spread of the Covid-19 virus since the end of 2019 in Indonesia has resulted in the Indonesian government taking several actions in various sectors. One of the government's efforts to handle and monitor the condition of the Covid-19 pandemic using information technology is by launching the PeduliLindungi application. With this application the government can monitor public data related to vaccination, tracing, telemedicine, and looking for rooms at the nearest hospital. The launch of the application and the obligation to use the PeduliLindungi application has received a response from the public, this can be monitored/seen from reviews on social media, news and also reviews on the Google Play Store regarding the application. Reviews from users on the Google Play Store can be used as parameters for input or feedback. This data is quite a lot and requires a long time to process it, even though the existing reviews could be useful as input for criticism and suggestions in future application development. From user review data, a classification process can be carried out based on sentiment type. Sentiment analysis is a branch of text classification which aims to classify sentiment (opinion) whether the text contains negative opinions, positive or neutral opinions. The aim of this research is to apply sentiment analysis to user review data of the PeduLindungi application into positive and negative classes using the Naive Bayes Classifier and Support Vector Machine classification algorithms. The dataset used after going through data pre-processing was 10,616 records. The results of testing and model evaluation carried out by randomly dividing training data and testing data obtained accuracy values, for the Naive Bayes Classifier method it was 77% and the Support Vector Machine had higher accuracy, namely 81%.