The Implementation of Probabilistic Neural Networks to Sentiment Analysis of National Principle and Religion Issues in Indonesia

Abstract

Indonesia's diverse population presents critical issues, including political, social, religious freedom, National Principle, and conflict issues. With the widespread use of internet and social media, machine learning technology has become a tool to analyze public opinions about anti-Pancasila, intolerance, khilafah, and radicalism. This study examines the frequency of discussions on these topics on social media and utilizes probabilistic neural networks (PNN) to classify text data. The study conducted several trials with different parameters and employed hold-out and K-fold cross-validation schemes. The results show that the radicalism keyword had the best accuracy of 64.9% on the hold-out with a spread of 0.1 and 20% of testing data, while the keyword khilafah had the best accuracy of 87.9% with a spread of 0.001 and K=6 in the K-fold cross-validation. The study also finds that cross-validation has better accuracy than hold-out to analyze the distribution of data. In 2019, the sentiment analysis revealed that almost 50% of Indonesians show neutral attitudes toward these four critical issues. This study demonstrates the potential of machine learning technology to analyze public opinions on complex sociopolitical issues in Indonesia.