Tweet Data Research Articles

BERT-based language model for accurate drug adverse event extraction from social media: implementation, evaluation, and contributions to pharmacovigilance practices.

Social media platforms serve as a valuable resource for users to share health-related information, aiding in the monitoring of adverse events linked to medications and treatments in drug safety surveillance. However, extracting drug-related adverse events accurately and efficiently from social media poses challenges in both natural language processing research and the pharmacovigilance domain. Recognizing the lack of detailed implementation and evaluation of Bidirectional Encoder Representations from Transformers (BERT)-based models for drug adverse event extraction on social media, we developed a BERT-based language model tailored to identifying drug adverse events in this context. Our model utilized publicly available labeled adverse event data from the ADE-Corpus-V2. Constructing the BERT-based model involved optimizing key hyperparameters, such as the number of training epochs, batch size, and learning rate. Through ten hold-out evaluations on ADE-Corpus-V2 data and external social media datasets, our model consistently demonstrated high accuracy in drug adverse event detection. The hold-out evaluations resulted in average F1 scores of 0.8575, 0.9049, and 0.9813 for detecting words of adverse events, words in adverse events, and words not in adverse events, respectively. External validation using human-labeled adverse event tweets data from SMM4H further substantiated the effectiveness of our model, yielding F1 scores 0.8127, 0.8068, and 0.9790 for detecting words of adverse events, words in adverse events, and words not in adverse events, respectively. This study not only showcases the effectiveness of BERT-based language models in accurately identifying drug-related adverse events in the dynamic landscape of social media data, but also addresses the need for the implementation of a comprehensive study design and evaluation. By doing so, we contribute to the advancement of pharmacovigilance practices and methodologies in the context of emerging information sources like social media.

Sentiment analysis of Indonesian tweets on COVID-19 and COVID-19 vaccinations

Background Sentiments and opinions regarding COVID-19 and the COVID-19 vaccination on Indonesian-language Twitter are scarcely reported in one comprehensive study, and thus were aimed at our study. We also analyzed fake news and facts, and Twitter engagement to understand people’s perceptions and beliefs that determine public health literacy. Methods We collected 3,489,367 tweets data from January 2020 to August 2021. We analyzed factual and fake news using the string comparison method. The difflib library was used to measure similarity. The user’s engagement was analyzed by averaging the engagement metrics of tweets, retweets, favorites, replies, and posts shared with sentiments and opinions regarding COVID-19 and COVID-19 vaccination. Result Positive sentiments on COVID-19 and COVID-19 vaccination dominated, however, the negative sentiments increased during the beginning of the implementation of restrictions on community activities (PPKM). The tweets were dominated by the importance of health protocols (washing hands, keeping distance, and wearing masks). Several types of vaccines were on top of the word count in the vaccine subtopic. Acceptance of the vaccination increased during the studied period, and the fake news was overweighed by the facts. The tweets were dynamic and showed that the engaged topics were changed from the nature of COVID-19 to the vaccination and virus mutation which peaked in the early and middle terms of 2021. The public sentiment and engagement were shifted from hesitancy to anxiety towards the safety and effectiveness of the vaccines, whilst changed again into wariness on an uprising of the delta variant. Conclusion Understanding public sentiment and opinion can help policymakers to plan the best strategy to cope with the pandemic. Positive sentiments and fact-based opinions on COVID-19, and COVID-19 vaccination had been shown predominantly. However, sufficient health literacy levels could yet be predicted and sought for further study.

Malicious Social Bot Using Twitter Network Analysis in Django

Malicious social bots generate fake tweets and automate their social relationships either by pretending to be a followers or by creating multiple fake accounts with malicious activities. Moreover, malicious social bots post shortened malicious URLs in the tweets to redirect the requests of online social networking participants to some malicious servers. Hence, distinguishing malicious social bots from legitimate users is one of the most important tasks in the Twitter network. To detect malicious social bots, extracting URL-based features (such as URL redirection, frequency of shared URLs, and spam content in URL) consumes less amount of time in comparison with social graph-based features (which rely on the social interactions of users). Furthermore, malicious social bots cannot easily manipulate URL redirection chains. In this article, learning automata-based malicious social bot detection (LA-MSBD) algorithm is proposed by integrating a trust computation model with URL-based features for identifying trustworthy participants (users) in the Twitter network. The proposed trust computation model contains two parameters, namely, direct trust and indirect trust. Moreover, the direct trust is derived from Bayes’ theorem, and the indirect trust is derived from the Dempster– Shafer theory (DST) to determine the trustworthiness of each participant accurately. Finally, we shown the user tweet data in terms of graph visualization of bar chart and pie chart of the system. Experimental results shown the better performance of the system.

Analisis Opini Terhadap Aplikasi Riliv di Twitter Menggunakan Algoritma Naïve Bayes dan Random Forest

In the current era of technological advancement and the internet, people can easily access various information. This technological advancement brings innovation in the mental health field, such as services in the form of apps. This research conducts sentiment analysis using the Naïve Bayes and Random Forest algorithms. The study aims to analyze Twitter users’ opinions regarding the Riliv apps and compare the results of classification using Naïve Bayes and Random Forest. This research methodology uses the AI Project Cycle method. The data used is tweet data from Twitter with the keyword 'aplikasi riliv’. The dataset consisted of 1035 data, which was processed to produce 273 positive, 273 neutral, and 39 negative sentiments data. The Naïve Bayes and Random Forest algorithms were applied to compare the classification results of the two. The most optimal classification results are Naïve Bayes with SMOTE with the division of 90% training data and 10% testing data, which results in an accuracy value of 82.72%, a value of precision is 82.89% and a value of recall is 82.72%. Based on the results of the distribution of sentiment data, most users gave positive reviews and were knowledgeable about the Riliv application, while only a few were disappointed

Pengembangan Aplikasi Analisis Sentimen Penerapan PPKM Level 3

This study aims to analyze tweets about the imposition of Community Activity Restrictions (PPKM) level 3 to find the number of positive, negative, and neutral responses by building webbased software to conduct sentiment analysis on tweets data with the keywords “PPKM level 3". This research consists of two increments. The first increment is modeling and sentiment analysis and the second increment is making a website for visualizing the results. The results in the first increment are a model to perform sentiment analysis with an accuracy of 76.59% for training and for testing, with an accuracy of f1-score 75% with the test results on the macro average precision 82%, recall 73%, f1-score 75% with support of 20 and weighted average precision 80%, recall 75%, f1-score 75% with support of 20 and dataset analysis results with 30.82% positive sentiments, 17.80% neutral sentiments, and 51.36% negative sentiments. The result in the second increment is a website for visualization.

Sentiment analysis of Indonesian tweets on COVID-19 and COVID-19 vaccinations.

Sentiments and opinions regarding COVID-19 and the COVID-19 vaccination on Indonesian-language Twitter are scarcely reported in one comprehensive study, and thus were aimed at our study. We also analyzed fake news and facts, and Twitter engagement to understand people's perceptions and beliefs that determine public health literacy. We collected 3,489,367 tweets data from January 2020 to August 2021. We analyzed factual and fake news using the string comparison method. The difflib library was used to measure similarity. The user's engagement was analyzed by averaging the engagement metrics of tweets, retweets, favorites, replies, and posts shared with sentiments and opinions regarding COVID-19 and COVID-19 vaccination. Positive sentiments on COVID-19 and COVID-19 vaccination dominated, however, the negative sentiments increased during the beginning of the implementation of restrictions on community activities (PPKM). The tweets were dominated by the importance of health protocols (washing hands, keeping distance, and wearing masks). Several types of vaccines were on top of the word count in the vaccine subtopic. Acceptance of the vaccination increased during the studied period, and the fake news was overweighed by the facts. The tweets were dynamic and showed that the engaged topics were changed from the nature of COVID-19 to the vaccination and virus mutation which peaked in the early and middle terms of 2021. The public sentiment and engagement were shifted from hesitancy to anxiety towards the safety and effectiveness of the vaccines, whilst changed again into wariness on an uprising of the delta variant. Understanding public sentiment and opinion can help policymakers to plan the best strategy to cope with the pandemic. Positive sentiments and fact-based opinions on COVID-19, and COVID-19 vaccination had been shown predominantly. However, sufficient health literacy levels could yet be predicted and sought for further study.

Application of Social Network Analysis for Comparison and Ranking of Internet Service Providers

In this digital era, the Internet has become a basic necessity in life. This has had a significant impact on the growth of internet service provider (ISP) companies in Indonesia. Comparison and ranking of ISPs is needed to make it easier for users to choose services according to their needs as well as to encourage healthy competition between ISPs in improving their services. The problem is ranking ISPs using conventional methods (surveys) to obtain primary data is expensive and takes a long time. On the other hand, Social Network Analysis (SNA) is a method that has been widely used to understand customer desires by extracting information from social media. This information is in the form of User Generated Content (UGC), namely track records left by customers on social media. This research aims to measure the ISP rankings of Indihome, Biznet and FirstMedia using UGC data. The research method used is to collect consumer tweet data rapidly, carry out preprocessing to eliminate irrelevant data and apply SNA, including network structure analysis in the form of visualization and network property analysis with the Gephi application, as well as network content analysis in the form of sentiment analysis and WordCloud analysis. The number of dominant network properties and sentiment analysis calculates ISP ranking. Apart from that, the results of this SNA are in the form of recommendations for ISPs to improve services to customers.

Analysis of the hikikomori phenomenon - an international infodemiology study of Twitter data in Portuguese.

Hikikomori refers to the extreme isolation of individuals in their own homes, lasting at least six months. In recent years social isolation has become an important clinical, social, and public health problem, with increased awareness of hikikomori around the globe. Portuguese is one of the six most spoken languages in the world, but no studies have analysed the content regarding this phenomenon expressed in Portuguese. To explore the hikikomori phenomenon on Twitter in Portuguese, utilising a mixed-methods approach encompassing content analysis, emotional analysis, and correlation analysis. A mixed methods analysis of all publicly available tweets in thePortuguese language using a specific keyword (hikikomori) between 1st January 2008 and 19th October 2022. The content analysis involved categorising tweets based on tone, content, and user types, while correlation analysis was used to investigate user engagement and geographical distribution. Statistical analysis and artificial intelligence were employed to classify and interpret the tweet data. Among the total of 13,915 tweets generated, in terms of tone 10,731 were classified as "negative", and 3184 as "positive". Regarding content, "curiosities" was the most posted, as well as the most retweeted and liked topic. Worldwide, most of the hikikomori related tweets in Portuguese were posted in Europe, while "individuals with hikikomori" were the users most active posting. Regarding emotion analysis, the majority of tweets were "neutral". These findings show the global prevalence of the discourse on hikikomori phenomenon among Portuguese speakers. It also indicates an increase in the number of tweets on this topic in certain continents over the years. These findings can contribute to developing specific interventions, support networks, and awareness-raising campaigns for affected individuals.

Twitter Sentiment Analysis in Tourism with Polynomial Naïve Bayes Classifier

Lombok has become a favorited tourist destination in the world. Therefore, tourism is a mainstay sector in regional development in West Nusa Tenggara. The contribution of the tourism sector shows an increasing trend. Tourist expenditures are distributed to various sectors. The tourism sector has a positive impact on the regional economy. The local government has prepared to improve the quality and quantity of tourism in Lombok. The results of local government efforts need to be analyzed so that future policies are on target. Analysis can be done on the satisfaction of tourists who travel to Lombok. It would be very difficult to get satisfaction data from all tourists through questionnaires. But on the other hand, tourist satisfaction is usually posted on their social networks. One of the social media that is widely used by tourists is Twitter. Their tweets contain not only expressions of natural beauty but also criticism, suggestions, and complaints during their visit. In addition, the tweet data on twitter is open access. This study tries to analyze the sentiment on Twitter which contains tweets of tourists who have visited Lombok. Sentiment analysis is performed using the Polynomial Naive Bayes Classifier. Sentiment results are classified into positive and negative sentiments. The results of this sentiment are expected to help related agencies or other tourism actors to improve the quality and quantity of regional tourism. The results showed that the positive sentiment on the security factor were 50.65%, the cost 75.32%, accommodation 62.33% and the cleanness factor 77.92%.

Understanding sentiment toward racial unrest through temporal and geographic lenses: a multilevel-analysis across metropolitan areas in the United States

Racial unrest has long been a salient social issue in the United States. Time and space provide essential contexts for the emergence and evolution of racial unrest. However, the relationships between these contextual factors and public responses to racial unrest remain insufficiently explored. This study seeks to fill that gap, blending geocoded, time-stamped racial unrest tweet data with census information. It aims to explore how temporal elements and geographical characteristics of metropolitan areas contribute to the emergence of negative sentiment reactions to racial unrest on social media platforms. The racially charged unrest that transpired in Ferguson, Missouri in 2014 serves as our case study. We select 33 metropolitan regions across the U.S. for our analysis. Our findings indicate that temporal processes, encompassing circadian rhythms, weekday-weekend variations, and temporal decay, correlate with expressions of anxiety and anger, albeit not sadness. Furthermore, our analysis reveals geographical characteristics—notably income inequality and segregation, combined with the number of Black victims—to be associated with manifestations of anxiety.

Sentiment analysis of tweets using text and graph multi-views learning

With the surge of deep learning framework, various studies have attempted to address the challenges of sentiment analysis of tweets (data sparsity, under-specificity, noise, and multilingual content) through text and network-based representation learning approaches. However, limited studies on combining the benefits of textual and structural (graph) representations for sentiment analysis of tweets have been carried out. This study proposes a multi-view learning framework (end-to-end and ensemble-based) that leverages both text-based and graph-based representation learning approaches to enrich the tweet representation for sentiment classification. The efficacy of the proposed framework is evaluated over three datasets using suitable baseline counterparts. From various experimental studies, it is observed that combining both textual and structural views can achieve better performance of sentiment classification tasks than its counterparts.

Exploring Discussions About Virtual Reality on Twitter to Inform Brain Injury Rehabilitation: Content and Network Analysis.

Virtual reality (VR) use in brain injury rehabilitation is emerging. Recommendations for VR development in this field encourage end user engagement to determine the benefits and challenges of VR use; however, existing literature on this topic is limited. Data from social networking sites such as Twitter may further inform development and clinical practice related to the use of VR in brain injury rehabilitation. This study collected and analyzed VR-related tweets to (1) explore the VR tweeting community to determine topics of conversation and network connections, (2) understand user opinions and experiences of VR, and (3) identify tweets related to VR use in health care and brain injury rehabilitation. Publicly available tweets containing the hashtags #virtualreality and #VR were collected up to twice weekly during a 6-week period from July 2020 to August 2020 using NCapture (QSR International). The included tweets were analyzed using mixed methods. All tweets were coded using inductive content analysis. Relevant tweets (ie, coded as "VR in health care" or "talking about VR") were further analyzed using Dann's content coding. The biographies of users who sent relevant tweets were examined descriptively. Tweet data networks were visualized using Gephi computational analysis. A total of 260,715 tweets were collected, and 70,051 (26.87%) were analyzed following eligibility screening. The sample comprised 33.68% (23,596/70,051) original tweets and 66.32% (46,455/70,051) retweets. Content analysis generated 10 main categories of original tweets related to VR (ie, advertising and promotion, VR content, talking about VR, VR news, general technology, VR industry, VR live streams, VR in health care, VR events, and VR community). Approximately 4.48% (1056/23,596) of original tweets were related to VR use in health care, whereas 0.19% (45/23,596) referred to VR in brain injury rehabilitation. In total, 14.86% (3506/23,596) of original tweets featured commentary on user opinions and experiences of VR applications, equipment, and software. The VR tweeting community comprised a large network of 26,001 unique Twitter users. Users that posted tweets related to "VR in health care" (2124/26,001, 8.17%) did not form an interconnected VR network, whereas many users "talking about VR" (3752/26,001, 14.43%) were connected within a central network. This study provides valuable data on community-based experiences and opinions related to VR. Tweets showcased various VR applications, including in health care, and identified important user-based considerations that can be used to inform VR use in brain injury rehabilitation (eg, technical design, accessibility, and VR sickness). Limited discussions and small user networks related to VR in brain injury rehabilitation reflect the paucity of literature on this topic and the potential underuse of this technology. These findings emphasize that further research is required to understand the specific needs and perspectives of people with brain injuries and clinicians regarding VR use in rehabilitation.

Islamic Crowdfunding: A Twitter Sentiment Analysis

This research aims to measure public sentiment related to Islamic crowdfunding on the Twitter social media platform. The research method involves the extraction and classification of tweet data using a Python Library called VADER (Valence Aware Dictionary and Sentiment Reasoner). The research utilized tweet data posted in the past one year. The results showed fluctuations and decreases in the number of tweets discussing Islamic crowdfunding. Word cloud analysis on keywords shows that in positive sentiments, words such as "crowdfunding platform," "crowdfunding impact," "crowdfunding," "Islamic finance," and "inclusivity Islamic finance" dominate the visualization. Overall, the majority of tweets expressed a positive attitude towards Islamic crowdfunding, with 44.0% positive sentiment. A total of 47.2% of tweets showed neutral sentiment, while 8.7% showed negative sentiment. These results illustrate that people generally give positive support to the concept of Islamic crowdfunding, although there is still a small proportion of tweets that express less favorable views. This research provides valuable insights into people's perceptions and responses to Islamic crowdfunding in cyberspace.

Six classes named entity recognition for mapping location of Indonesia natural disasters from twitter data

PurposeThe purpose of this study is to provide the location of natural disasters that are poured into maps by extracting Twitter data. The Twitter text is extracted by using named entity recognition (NER) with six classes hierarchy location in Indonesia. Moreover, the tweet then is classified into eight classes of natural disasters using the support vector machine (SVM). Overall, the system is able to classify tweet and mapping the position of the content tweet.Design/methodology/approachThis research builds a model to map the geolocation of tweet data using NER. This research uses six classes of NER which is based on region Indonesia. This data is then classified into eight classes of natural disasters using the SVM.FindingsExperiment results demonstrate that the proposed NER with six special classes based on the regional level in Indonesia is able to map the location of the disaster based on data Twitter. The results also show good performance in geocoding such as match rate, match score and match type. Moreover, with SVM, this study can also classify tweet into eight classes of types of natural disasters specifically for the Indonesian region, which originate from the tweets collected.Research limitations/implicationsThis study implements in Indonesia region.Originality/value(a)NER with six classes is used to create a location classification model with StanfordNER and ArcGIS tools. The use of six location classes is based on the Indonesia regional which has the large area. Hence, it has many levels in its regional location, such as province, district/city, sub-district, village, road and place names. (b) SVM is used to classify natural disasters. Classification of types of natural disasters is divided into eight: floods, earthquakes, landslides, tsunamis, hurricanes, forest fires, droughts and volcanic eruptions.

Analyzing Twitter Users’ Sentiments on the Surge of Fuel Oil Prices in Indonesia using the K-Nearest Neighbor Algorithm

Sentiment analysis offers an effective solution for automating the classification of text data based on polarity, facilitating the assessment of public opinion. Among various social media platforms, Twitter stands out as a significant source of concise textual data reflecting users’ viewpoints on diverse topics. Notably, the recent surge in the price of fuel oil (BBM) in Indonesia has sparked considerable discussion and expression on Twitter. In this study, our objective was to perform a comprehensive sentiment analysis of Twitter users’ reactions to the rising fuel prices in Indonesia by employing the K-Nearest Neighbor (K-NN) algorithm. The research followed a structured approach encompassing data collection, text preprocessing, data labeling, feature extraction, data splitting, classification, and algorithm performance evaluation. The results revealed a dominance of negative sentiments among the 5,000 collected tweet data. The sentiments were categorized as 54.6% negative, 31.8% positive, and 13.6% neutral. This indicates a prevailing level of dissatisfaction and concern expressed by Twitter users regarding the fuel price increase. The K-NN algorithm’s classification performance was most promising when evaluated in an 80:20 data ratio experiment, yielding an accuracy rate of 65%, precision of 74%, recall of 45%, and an error rate of 35%. These findings suggest that the K-NN algorithm is valuable for effectively gauging public sentiment towards the escalating fuel prices in Indonesia. This research highlights the potential of sentiment analysis and the K-NN algorithm in assessing public reactions to significant events, providing valuable insights for policymakers and stakeholders in the energy sector.

Data tweet clustering using bidirectional gated recurrent unit and k-prototype for the Indonesian political year

As time passes, social media, which was formerly used as a means of communication between users, is experiencing a transition as a means for broadcasting information, conducting business, advertising, and even political campaigning. In elections, social media is also used to discredit political opponents to reduce the electability of opposing candidate. Spreading hate speech and fake news to undermine the electability of opposing candidate is a common violation of the law committed by supporters of one candidate over another. Considering that the number of social media users increases annually at a very rapid rate, the hazard of social media abuse has the potential to grow. In 2022, Indonesia had 191 million social media users in January 2022. Obviously, this will make the election situation more tumultuous and has the potential to cause societal divisions. The government must have a control system in place to screen social media content that can be considered illegal. In this study, fake news and hate speech are classified using the Bidirectional Gated Recurrent Unit (BiGRU). Lastly, K-Prototype was used to do clustering based on categorization dimensions and probable distribution to identify which clusters had the greatest risk of breaking the law, creating confusion, and dispersing broadly throughout society. It is hoped that the clusters that are created will represent the levels of priority of tweet data that requires prompt attention from the government to prevent it from spreading and inciting social unrest. Based on the results of the analysis, the BiGRU fake news model yields a F1-score of 95%, while the BiGRU hate speech model yields a F1-score of 90%. Clustering data using K-Prototype in this research can reduce the number of tweet data from 13,183 to 1,791 data. These new data are considered as a priority that must be pursued in preventing social media disputes.

Gerakan Opini Digital pada #TolakUUCiptaKerja di Twitter dengan Pendekatan Social Network Analysis

The ratification of the UU Cipta Kerja led to objections that were echoed through #TolakUUCiptaKerja whose opinions were widely spread on Twitter. The purpose of this study was to determine the digital opinion movement #TolakUUCiptaKerja on Twitter. The research method used was a mix method by combining quantitative communication network statistics with a sample of 1000 tweet data, the number of actors 355 and 344 relations using netlytic and gephi. Qualitative to analyze text with a digital movement of opinion that results in the dissemination of opinions on the structure of the communication network. The results showed that the #TolakUUCiptaKerja movement was able to create mobility in the opinion of Twitter users, namely 752 or 75.2% positive reactions, 7 or 7% negative reactions and 241 or 24.1% irrelevant in communication networks. The positive reaction was dominated by netizen support for rejecting the legalization of the UU Cipta Kerja. The spread of this hashtag was assisted by popular actors, namely @ramlirizal and @ sandalista1789 as important actors, and 205 out of 355 actors were important actors

Analysis of Twitter User Sentiment on Presidential Candidate Anies Baswedan Using Naïve Bayes Algorithm

Indonesian hold presidential election in 2024. One of the most discussed topics by public is the presidential candidates. The discussion about the presidential candidate certainly reaped various kinds of responses from public, ranging from support to statements of disapproval. This research was limited to the candidacy of Anies Baswedan as a presidential candidate before a vice president candidate as his pair was selected. The purpose of this study is to conduct a sentiment analysis of public responses regarding Indonesia 2024 presidential candidate Anies Baswedan using tweets data from October 2022 to January 2023 using the naïve bayes classifier algorithm. This is expected to provide an overview of the public opinions on Twitter. Three test models were carried out with differences in the division of the amount of training data and test data, respectively 60%:40%, 70%:30% and 80%:20%. The test results showed the highest accuracy level was obtained by the 3rd model using training and testing data of 80%:20% with an accuracy value of 76.21%. Further research is recommended to conduct sentiment analysis on the pairs of Presidential and Vice-Presidential candidates who have been officially registered with the General Election Commission using various other classification algorithms.

Mining Netizen's Opinion on Sustainable Agriculture: Sentiment Analysis of Twitter Data

This research aims to measure public sentiment related to sustainable agriculture on the Twitter social media platform. The research method involves the extraction and classification of tweet data using a Python Library called VADER (Valence Aware Dictionary and Sentiment Reasoner). The research utilized tweet data posted in the past one year. The results showed fluctuations and decreases in the number of tweets discussing sustainable agriculture. The location with the most tweet activity around sustainable agriculture was Brussels, Belgium, with 642 tweets during the observation period. Word cloud analysis on keywords showed that in positive sentiments, words such as "food security" and "climate change" dominated the visualization. On the other hand, in negative sentiments, words such as "farmer" and "private farmland" appeared more frequently. Overall, the majority of tweets expressed a positive attitude towards sustainable agriculture, with 68.5% positive sentiment. A total of 22.3% of tweets showed neutral sentiments, with no strong positive or negative tendencies. Only 9.1% of tweets contained negative sentiment, indicating that a small proportion of tweets expressed less favorable views towards sustainable agriculture.

K-Means Clustering to Identity Twitter Build Operate Transfer (BOT) on Influential Accounts

Twitter is a popular social media with hundreds of millions of users, but some are not human. About 48 million accounts are created by Build Operate Transfer (BOT), which represents up to 15% of all accounts. BOTs are created for various purposes, one of which is to post information about news automatically. However, BOTs have also been abused, such as spreading hoaxes or influencing public perception of a topic. The research aimed to determine which Twitter accounts were identified as BOT accounts based on predefined attributes. The research used tweet data from 213 Twitter accounts. The accounts used as test data were accounts that had influence. After that, the data were clustered using k-means using the attributes of retweets + replies count, followers count, account age, friends count, status count, digits count in name, username length, name similarity, name ratio, and likes count. The results show the optimal number of clustering at k = 3 on the Sum of Squared Errors (SSE) evaluation and the Elbow method and the best quality and cluster power at k = 2 on the silhouette coefficient. It shows that the clustered accounts with the highest number of members on each attribute are places for accounts with high BOT scores from several aspects of the BOT score type.