Analyzing Public Trust in Presidential Election Surveys: A Study Using SVM and Logistic Regression on Social Media Comments

Abstract

In the context of democracy in Indonesia, elections play a crucial role, and survey agencies often publish their results on social media. User responses, especially from voters, often express dissatisfaction, including distrust, insults, and negative comments, if the candidate they support receives low survey results. Therefore, this study aims to examine the level of public trust in the survey results of Presidential candidates in 2024 using the Support Vector Machine (SVM) and Logistic Regression algorithms. The study utilized data from 1778 Instagram comments and 985 Twitter tweets. The process involved problem identification, data collection, and system implementation, such as preprocessing, labeling, SMOTE, TF-IDF, data splitting, model classification, and evaluation. The results show that SVM with an 80% training data and 20% test data scenario provides high accuracy, namely 93.19% from Instagram and 91.19% from Twitter. Logistic Regression, with the highest accuracy of 89.79% from Instagram and 88.01% from Twitter in the same scenario. Sentiment analysis using SVM scenario one resulted in 195 positive comments and 216 negative comments. Logistic Regression scenario one shows 180 positive sentiments and 216 negative sentiments. From the classification results, it can be concluded that the level of public trust tends to be negative towards the survey results of the 2024 Presidential candidates, both using SVM and Logistic Regression.