PENGIMPLMENTASIAN ALGORITMA LONG SHORT-TERM MEMORY UNTUK MENDETEKSI UJARAN KEBENCIAN PADA APLIKASI TWITTER

Entering 2022, the number of internet users in Indonesia has reached 204.7 million users, where most of the internet us-age is for social media. Along with the high number of social media users, the Directorate of Cyber Crime of the Criminal Investigation Unit of the National Police found 89 verified social media content containing hate speech during the Febru-ary-March 2021 period, where the most content came from the Twitter application. Therefore, a research was conducted by implementing Machine learning to detect hate speech on the Twitter application using the Long short-term memory meth-od. Twitter data absorption was carried out by implementing the Tweepy Library by Muhammad Okky Ibrohim which was accessed via Kaggle for about 7 months, from March 20, 2018 to September 10, 2018. Data that has gone through text processing is then made into tokens which are a series of integer values. Then the LSTM model is built by compiling the input layer, LSTM layer, and output layer, to be trained later with training data that has been separated from the dataset. The researcher found that the results of the model training showed an accuracy of 95.74% and a loss value of 0.3463. When the trained model is used to make predictions on the test data, the researcher gets an accuracy value of 90% which indicates the model has made accurate predictions. Based on the model's performance in detecting hate speech, research-ers can conclude that hate speech detection on JIPI (Jurnal Ilmiah Penelitian dan Pembelajaran Informatika) Journal homepage: https://jurnal.stkippgritulungagung.ac.id/index.php/jipi ISSN: 2540-8984 Vol. 8, No. 2, Juni 2023, Pp. 551-560 552 Pengimplmentasian Algoritma Long Short-Term Memory Untuk Mendeteksi Ujaran Kebencian Pada Aplikasi Twitter Twitter can be done using Machine learning and Long short-term memory (LSTM) algorithms with a fairly high level of accuracy.