Analysis of Public Sentiment Towards POLRI's Performance using Naive Bayes and K-Nearest Neighbors

Abstract

Using Twitter as a platform for sharing information includes tracking public perceptions of the performance of the Indonesian National Police (POLRI). Public sentiment assists as a gauge for evaluating POLRI's operational capabilities and supports decision-making processes to enhance the organization's reputation. However, raw public opinion data often requires careful analysis for decision-making. Hence, conducting sentiment analysis of Twitter data is crucial. This analytical process involves extracting and classifying opinions into neutral, positive, and negative sentiments. This study employs two distinct sentiment analysis methods: the Naive Bayes algorithm and the K-Nearest Neighbors. Analysis of 1285 tweets reveals prevailing satisfaction with POLRI's performance, indicated by many positive sentiments. However, there is also a notable number of negative feelings. The assessment from confusion matrix results demonstrate that the Naive Bayes algorithm achieves 99.03% accuracy, while the K-Nearest Neighbors algorithm achieves 95.33% accuracy. By leveraging insights from public opinion data, POLRI can make more accurate and timely decisions, enabling it to better fulfill the community's needs and expectations. This strategic use of data enhances service quality and bolsters POLRI's favorable image among the public fosters more harmonious relationships and enhances public trust in law enforcement agencies.