Sentiment Analysis on Social Media X (Twitter) Against ChatGBT Using the K-Nearest Neighbors Algorithm

E-ISSN : 2807-9035 Volume 4, Number 1, May 2024 https://doi.org/10.47709/brilliance.v4i1.4105 Sentiment Analysis on Social Media X (Twitter) Against ChatGBT Using the K-Nearest Neighbors Algorithm Asep Arwan Sulaeman1*, Muhtajuddin Danny2, Sufajar Butsianto3, Suria Pratama4 1,2,3,4 1 Informatics Engineering Study Program, Faculty of Engineering, Pelita Bangsa University, Indonesia , , , *Corresponding Author Article History: Submitted: 15-06-2024 Accepted: 16-06-2024 Published: 01-07-2024 Keywords: ChatGPT; K-Nearest Neighbors (KNN); sentiment analysis; Twitter. Brilliance: Research of Artificial Intelligence is licensed under a Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). ABSTRACT This research aims to analyze the public's response to ChatGPT through data obtained from Twitter. Apart from that, it is also to understand whether people's responses tend to be positive or negative towards ChatGPT, as well as to test the performance of the K-Nearest Neighbors (KNN) method in classifying sentiment patterns in tweet data. The sentiment analysis method is carried out by dividing public responses into positive and negative categories. Next, the performance of the K-Nearest Neighbors (KNN) method was tested with varying k values to classify sentiment patterns in tweet data. This testing includes dataset division, vectorization of text data using TF-IDF, initialization and training of the KNN model, and evaluation of model performance using metrics such as precision, recall, and f1-score. The results of sentiment analysis show that the majority of people's responses to ChatGPT are positive (74.3%), while 25.7% of responses are negative. Performance testing of the KNN model shows that the highest accuracy of 88% is achieved when the k value is 5. Evaluation of model performance also shows satisfactory levels of precision, recall and f1-score. Based on the research results, it was concluded that sentiment analysis and classification using KNN were effective in understanding people's responses to ChatGPT. INTRODUCTION In a situation of increasingly rapid advances in information and communication technology (Triyono & Febriani, 2018), ChatGPT has become an innovation that has attracted attention from the public. ChatGPT is a natural language processing technology or what is called (natural language processing / NLP) which is capable of receiving commands from users and can also answer user questions in the form of text or what can be called prompts that are entered into the application . Sentiment analysis is a method of extracting information from text which aims to understand whether an opinion or assessment from internet users on social media is positive, neutral or negative. This technique is used to evaluate personal opinions expressed by users, providing insight into their views on something (Fransiska Vina Sari & Arief Wibowo, 2019). Sentiment analysis can provide valuable insight into how users evaluate ChatGPT's performance and help identify potential improvements (Lubis et al., 2024). By continuously monitoring user sentiment, the development team can respond to feedback more effectively, ensuring that ChatGPT continues to evolve according to user expectations and needs (Irwansyah Suwahyu et al., 2024). This research uses the KNN classification method, namely K-Nearest Neighbors, this method was chosen because it has the advantage of classifying data based on the level of similarity with the nearest neighbors. KNN as a classification method is classified as a relatively easy approach without requiring complicated calculations (Endang Sholihatin et al., 2023). By using this method, it is hoped that this method can identify sentiment patterns that appear in tweets on Twitter (Novianti & Wibowo, 2022). The existence of ChatGPT as a smart language processing technology that is being widely discussed in various places has given rise to various views, both positive and negative, from users. Therefore, this study aims to investigate what people think about ChatGPT, whether their views tend to be positive or negative. In this research, the author will use a simple approach with the K-Nearest Neighbors (KNN) Method to better understand how people see the advantages and disadvantages of ChatGPT. Thus, this research not only helps understand people's overall opinion of ChatGPT, but also contributes to further understanding of how language processing technology is received by society (Irsalinda et al., 2021). LITERATURE REVIEW Faiza Rizqi Irawan, Ahmad Jazuli, Tutik Khotimah (Rizqi Irawan, 2022), in Sentiment Analysis Of Gojek Users Using The K-Nearset Neighbors Method explained that the application of the K-Nearest Neighbor method in classifying Twitter user responses can be a basis for evaluating and assessing Gojek services for the company. Testing this method using a confusion matrix on a dataset of 1409 shows an accuracy level of 79.43% with a value of k=15. This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 265 E-ISSN : 2807-9035 Volume 4, Number 1, May 2024 https://doi.org/10.47709/brilliance.v4i1.4105 The research process begins with collecting data (crawling data), managing data (data preprocessing), labeling the data (labeling), classifying using the KNN algorithm, and finally, carrying out evaluation. The evaluation results show an accuracy level of 94.33% in classifying data. This opinion is the result of research by Muhamad Trian Diwandanu, Lu'lu Mawaddah Wisudawati (Diwandanu & Wisudawati, 2023), in his journal entitled Sentiment Analysis of Twit Maxim On Twitter Using R Programming And K Nearest Neighbors. The Flouting Maxim on Twitter Influencers’ Tweets, this research aims to determine the use of maxim principles in tweets made by certain social media influencers in Indonesia whose method was carried out qualitatively. This research is limited to whether users comply with cooperative principles, maxims, especially the maxim of relevance, what purpose users usually violate these maxims(Hassani, 2019). The results obtained vary: most of the conversations do not meet the principle of the maxim of relevance, or in other words do not imply the principle of the maxim of relevance. Moreover, the goal is to crack jokes, and to keep the conversation going smoothly while engaging in good manners (Syaifuddin et al., 2021). Analysis Of Sentiment Towards The Community For New Banknote Using The K-Nearest Neighbor (KNN) Algorithm is research from Septi Hasanah, Intan Purwasih, Imam Santoso (Septi Hasanah et al., 2023). The research used K-Nearest Neighbor on 510 data for sentiment analysis, achieving an accuracy of 75.06%. The process involves data crawling, pre-processing, and classification with RapidMiner. Evaluation was carried out by taking True Positives and True Negatives, showing positive sentiment of 67.74% and negative 76.65%. In conclusion, the (...truncated)