Analisis Sentimen Masyarakat Pada Twitter Terhadap Debt Collector Menggunakan Metode Naive Bayes Classifier

Abstract

Currently, many people need cash loans. Sourced from the results of the Financial Services Authority (OJK) report, the amount of online loan distribution has reached IDR 18.73 trillion throughout 2022. Debt collectors are present to collect for people who are unable to pay off the loans that have been submitted. Even so, the performance of debt collectors has experienced a lot of criticism from the public, the performance of debt collectors is often discussed on social media twitter. Therefore, it is necessary to analyze the issue of debt collectors on twitter to see public opinion about the practice of debt collectors themselves. The data used in this study amounted to 600 data. Data labeling uses Indonesian linguists as validators to determine the sentiment class. In the preprocessing stage the data is cleaned to reduce attributes that have little effect on the classification process. The highest accuracy result obtained using Naive Bayes Classifier and TF-IDF is 78.3% with a data percentage of 90:10 (90% training data and 10% test data). For the highest precision value obtained from testing with the use of 90:10 data using Naive Bayes Classifier with the BoW method, which is 80%. While the highest recall value is obtained from testing with the use of 90:10 data using the BoW method which is 80%. The use of BoW succeeded in increasing the accuracy value in most of the data sharing scenarios in the test. The classification process produces the most frequently occurring words in each sentiment class visualized with word clouds and fishbone diagrams. The word "chase" is the most dominant word in negative tweet data, while the word "help" is the most dominant in positive tweet data against debt collectors on twitter. The depiction using the fishbone diagram provides a solution to the negative opinions and experiences of the public towards debt collectors, one of which is the use of harsh words which can be overcome by conducting training for all team members on communication ethics.