Development of a Community Sentiment Analysis System for Chatgpt on Twitter Using a Comparison of the Naive Bayes Classifier and K-Nearest Neighbors Methods

Chatgpt merupakan chatbot yang memiliki model kecerdasan buatan dan dirancang untuk menghasilkan text bahasa  manusia. Chatgpt menggemparkan dunia sosial media dikarenakan dapat membantu tugas copywriter dengan banyak manfaat. Chatgpt juga dikhawatirkan membawa pengaruh negatif seperti alat pelanggaran akademik dalam ujian  online. Penelitian menggunakan media Twitter untuk memperoleh dataset,  Crawling  data dari  media Twitter menghasilkan 1229 data yang terdiri 629 sentimen positif 300 sentimen negatif dan 300 sentimen netral. Teknik pengujian menggunakan split data dan K-Fold Cross Validation. Hasil uji coba menunjukkan bahwa metode Naive Bayes lebih baik daripada K-Nearest Neighbors pada kedua pengujian. Pengujian  dengan  hasil akurasi tertinggi sebesar 82,25%, nilai presisi  81,91%, recall 82,25% dan f1-score 81,37% untuk metode Naive Bayes dengan pengujian split data. Di sisi lain metode K-Nearest  Neighbors dengan split data memiliki hasil akurasi 80,43%,  nilai  presisi 80,97%, recall 80,43% dan f1-score 80,33%. Metode Naive Bayes akurasinya  lebih  tinggi  1.82%  pada  pengujian  split  data dan lebih tinggi 4.57% untuk pengujian k-fold cross validation

Chatgpt is a chatbot that has an artificial intelligence model and is designed to produce human language text. Chatgpt is taking the world of social media by storm because it can help copywriters with many benefits. It is also feared that Chatgpt will have a negative influence, such as being a tool for academic violations in online exams. The research used Twitter media to obtain a dataset. Crawling data from Twitter media produced 1229 data consisting of 629 positive sentiments, 300 negative sentiments and 300 neutral sentiments. The testing technique uses split data and K-Fold Cross Validation. The test results show that the Naive Bayes method is better than K-Nearest Neighbors in both tests. The test with the highest accuracy results was 82.25%, precision value 81.91%, recall 82.25% and f1-score 81.37% for the Naive Bayes method with split data testing. On the other hand, the K-Nearest Neighbors method with split data has an accuracy of 80.43%, a precision value of 80.97%, a recall of 80.43% and an f1-score of 80.33%. The Naive Bayes method has a higher accuracy of 1.82% for split data testing and 4.57% higher for k-fold cross validation testing