@binus.ac.id
Computer Science
Bina Nusantara University
Affective Computing, Social Signal Processing, Virtual Agents, Deep Learning, Game Technology
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Andry Chowanda, Nadia Nadia, and Lie Maximilianus Maria Kolbe
Institute of Advanced Engineering and Science
Several industries use clickbait techniques as their strategy to increase the number of readers for their news. Some news companies implement catchy headlines and images in their news article links, with the expectation that the readers will be interested in reading the news and click the provided link. The majority of the news is not hoax news. However, the content might not be as grand as the catchy headlines and images provided to the readers. This research aims to explore the classification model using machine learning to identify if the headlines are classified as clickbait in online news. This research explores several machine learning techniques to classify clickbait in online news and comprehensively explain the results. Several popular machine learning techniques were implemented and explored in this research. The results demonstrate that the model trained with fast large margin provides the best accuracy and classification error (90% and 10%, respectively). Moreover, to improve the performance, bidirectional encoder representations from transformers architecture was used to model clickbait in online news. The best BERT model achieved 98.86% in the test accuracy. BERT model requires more time to train (0.9 hour) compared to machine learning (0.4 hour).
Hansen Artajaya, Julieta, Jose Giancarlos, Jurike V. Moniaga, and Andry Chowanda
IEEE
Enrichment program is an important activities for student in BINUS University to develop their skills to face the real world. There are seven tracks that can be chosen by student. Student is given the freedom to choose, but it lead to a problem where student can't make a good decision and from the preceding enrichment program, some student doesn't satisfied with their choice. Hence that, proposed a solution to build a recommender system to help student choose a more suitable option with their personality and goals, without forcing the option. This research aims to build the most accurate recommender system from personality test result and soft skill test result to recommend an enrichment track. Based on the previous research, it can be determined that this recommender system is suitable to be built using supervised learning method because the dataset provided is labelled. Because the label is originated from a continuous values, it is decided to use a regression method. The experiment is done using scikit-learn to simplify the workflow without compromising the effectiveness of the algorithm. Based on the experiment, decision tree with seven max depth yields a model with minimum root mean squared error (RMSE) and maximum R2.
Jasen Wanardi Kusno, Rindy Claudia Setiawan, Irene Anindaputri Iswanto, Esther Widhi Andangsari, and Andry Chowanda
IEEE
Affective Computing is the study of systems that can recognize underlying human emotions. To be able to detect this, the systems usually have a sensor that captures the features of the input to evaluate it further. Based on this idea, we tried to explore several state-of-art machine learning methods and deep learning methods that are used to classify emotions. However, recognizing emotions is relatively a daunting task for an un-social computer. There are several techniques to model the emotions from text, some implement machine learning, others take advantage of the deep learning technology. Therefore, we evaluated several models of Machine Learning methods and recent Deep Learning methods to make a comparison with previous related works. The dataset used in this research was from social media. Recognizing emotions from social media provides some non-verbal insights for the readers. The experiment demonstrates that deep learning models outperform several previous results and machine learning models on emotion recognition tasks. The results demonstrate that the model trained with BERT achieved the accuracy of 93.75%. Moreover, Love and Joy class is relatively challenging to distinguish.
Steven, Victor C. Malik, Jeklin Harefa, Alexander, and Andry Chowanda
IEEE
Pandemic hits the world catastrophically, leading to the disruption on our everyday life including how students learn in the pandemic era. Most of the students nowadays have difficulties preparing the way to face the world of work. This research proposes a serious game design to prepare university students for the world of work by combining the Six Facets of Serious Game Design and Adam's Ernest Game Design. Six Facets of Serious Game Design is the design pattern for serious games (e.g., pedagogical games) [1]. Two systems (i.e. the serious game and the baseline system) were designed, deployed and evaluated to evaluate the effectiveness of the proposed serious game design. The serious game design was built on an Android mobile phone. Moreover, a simple simulation that has identical scenarios with the serious game was also built as the baseline in the evaluation process. Thirty-seven respondents who will be or have been entering the workplace for their first time participated in the evaluation process. A Wilcoxon rank test was applied to the data as the data does not come from a normal distribution. The results show that the proposed serious game design provides more effective results than the baseline. The user experiences' score provided by the serious game design system is statistically higher than the baseline system. Moreover, the serious game system increases the learners' motivation compared to the baseline system.
Katriel Serafina Widjaja, Carla Chika Alamo, Anderies, and Andry Chowanda
IEEE
Micro-expressions are the subtle and rapid movements of human facial expressions that could reveal a person’s true emotions, including emotions that people attempt to suppress, hide, or restrain. Many recent papers have researched facial expression recognition systems in video sequences using GRU models. However, they haven’t found a good relevance of micro-expression (ME) in detecting deceptive behaviors. In order to improve and contribute to the development of the system, we propose a micro-expression lie detection system with GRU’s hyperparameter optimization and explore its accuracy. FER-2013 is used for expression recognition learning and the dataset containing video clips of courtroom trials is used for deception detection learning. Several normalization techniques are done in the process. The CNN model with eight convolutional layers and three fully linked layers is used to train the facial expression recognition system. To improve its accuracy, multiple GRU parameter settings are employed. We used the model on the test dataset after training it. The outcome shows that it was 92.31% accurate. The confusion matrix predicts 12 out of 13 outcomes, with 100% accuracy on the deceptive class and 85% accuracy on the truthful class.
Nathan Jacky Lee, Muhammad Devin Nayottama A.P., Anderies, and Andry Chowanda
IEEE
YouTube is the world's largest video streaming platform, where videos appear in users” recommendation lists as thumbnail images and titles. Content creators compete for viewer attention, but as of now, what drives the popularity of YouTube videos is still a young and growing body of research, especially regarding users” pre-view behavior - what drives them to click a video they see. This research contributes to the field by exploring new content-independent visual attributes of thumbnails and titles entirely extractable by code. Several Python libraries and AI models are used to extract data such as image complexity, thumbnail text, sentiment, and faces from a dataset of 1600 videos. The extracted data is visualized to explore their relationship with video view count. This research unexpectedly finds that the studied attributes have negligible effect strength on a video's ability to attract views. Findings and possibilities regarding this outcome are discussed.
Ronald Sumichael Sunan, Samuel Christopher, Novandy Salim, Anderies, and Andry Chowanda
Elsevier BV
Anderies, Rendy Adidarma, Maximillian Lemuel Chanyassen, Alexander Imanuel, and Andry Chowanda
Elsevier BV
William Rusdyputra and Andry Chowanda
Elsevier BV
Anderies, Maevy Marvella, Nissa Adila Hakim, Priskilla Adriani Seciawanto, and Andry Chowanda
Elsevier BV
Louis Vincent Sanjaya, Herolistra Baskoroputro, and Andry Chowanda
IEEE
In the current digital era, safeguarding data security is essential. However, some companies still rely on outdated data security systems. This study aims to investigate the implementation of blockchain technology in enhancing the security of financial asset management and corporate transactions. Blockchain is a technology that provides a secure and transparent distributed database for storing and validating encrypted transactions. The Python programming language will be used for developing the solution. This research will involve a literature review on the basic concepts of blockchain, financial asset security, and corporate transactions. Subsequently, an application for financial asset management and corporate transaction system will be designed and implemented using the Python programming language, aided by GUI tools like Tkinter. The study reveals that the implementation of blockchain applications has successfully elevated the level of security in recording financial transaction data. Moreover, the research findings demonstrate that this application can also be effectively utilized by individuals without prior experience in blockchain systems.
Gabriel Theron, Timothy Liundi, Andry Chowanda, and Anderies
IEEE
Jason Hartanto, Sean Matthew Wijaya, Anderies, and Andry Chowanda
IEEE
Axel Jeremy Oei, Michael Rio Agustino Tan, Anderies, and Andry Chowanda
IEEE
The attendance management system is crucial in educational bodies. Over the years, several attendance methods have been implemented, such as doing roll calls. This type of manual attendance has many flaws and can be improved. Research done on improving attendance management systems has been conducted by many people. We also aspire to improve attendance management systems by creating an automated attendance management system. A method of using face recognition to improve the attendance management system has been proposed. The face recognition of the attendance management system utilizes deep learning as its base. We tested this system on several students and asked for their feedback. The experiment subjects were satisfied by our system, and the results gathered were satisfactory as the system can recognize the students and register their attendance accordingly. This type of attendance management system could serve as an alternative and improve the flaws of manual attendance management systems.
Jason Adiwijaya, Venansius Reynardi Tanaya, Anderies, and Andry Chowanda
IEEE
Andry Chowanda, Nadia, and Diana
IEEE
Emotions are essential to social interaction between interlocutors. It provides essential meaning to social interaction. It is important to recognise emotions along with verbal cues to fully understand the true meaning of a conversation in a social interaction setting. Several techniques can be applied to recognise emotions in the social interaction setting automatically. Emotions can be recognised and interpreted from the interlocutor's facial cues (e.g., facial expression recognition task) by using a sensor such as a camera. Recognising emotions from facial cues has several problems to be solved. This research aims to model robust emotion recognition from facial cues to recognise emotions (particularly from Asian people), as there is only limited research done in this area. This research aims to propose an emotion recognition model from facial cues by implementing Vision Transformers architecture. The dataset implemented in this research was explicitly performed by local (i.e., Indonesian) actors (The Indonesian Mixed Emotion Dataset - IMED). The results show that the proposed model can achieve the best testing accuracy of 100% with the best training loss of 0.0001, with 0.53 hours of training times for 50 epochs.
Pattrick Ritter, Devan Lucian, Anderies, and Andry Chowanda
IEEE
Deepfake technology has become a significant concern due to its ability to create highly realistic fake videos and images, leading to the potential deception of individuals. Detecting deepfakes has become a critical research area in computer vision and multimedia forensics. This paper presents a comparative analysis of deepfake detection models, focusing on evaluating their accuracy and robustness. Four CNN models, namely ResNet-152, MobilenetV3, Convnext Large, and EffecientNetB7, were implemented and trained using a custom dataset obtained from FaceForensics++. The models were evaluated based on training accuracy, average loss, and testing accuracy. An LSTM layer was also incorporated into each model's architecture to leverage sequential information. The results demonstrate varying performance among the models, with EfficientNet B7 achieving the highest testing accuracy of 75%. The findings of this study provide insights for future research in this critical area.
Andry Chowanda, Esther Widhi Andangsari, Violitta Yesmaya, Tin-Kai Chen, and Hsiao-Lin Fang
IEEE
Depression is a common mental health disorder. It can greatly affect our daily lives. Depressed people are generally prone to negative emotions. Hence, recognising a sign of depression is an important task. Several techniques are proposed to model automatic depression recognition from several modalities. However, there is a limited number of datasets and research done in a local language (i.e. Indonesian). Expressing thoughts and feelings are unique based on their backgrounds (e.g. race, religion and culture). Hence, fine-tuning the model to a local language or culture is also important. This research aims to build a model using deep learning to recognise depression signs from the text in the local language (i.e. Indonesian). Seven models are proposed in this research to model depression recognition from social media. The result illustrates that combining Bidirectional Long short-term memory with Bidirectional Encoder Representations from Transformers architecture can improve the performance of the model.
Yohan Muliono, Ford Lumban Gaol, Andry Chowanda, and Widodo Budiharto
IEEE
Fake news so called made-up news intended to cause misinformation is identified as Hoax. In Indonesia, hoaxes could not be ignored, fairly low literacy rate is the cause of hoaxes could spread quickly in Indonesia, to make it even worse, there are so many people would believe in hoaxes without any intention to make sure of the integrity of the news. even the media in Indonesia are competing with one another to provide content with provocative names in the hopes of increasing their rating and reaching as many people as possible. In the course of this research, a method for identifying authors on the basis of their writings will be developed to reduce the spread of hoaxes and encourage people to re-evaluate their decision to create harmful content and hoax news in the future. In this investigation, a transformer-based methodology focusing on IndoBERTbase, IndoBERTbase-uncased, and IndoBERTtweet will be employed in conjunction with the self-collected Indonesian tweet data that will be crawled from Indonesian Social Actor Focused on politicians in Twitter and will be combined with another topic in the future study. This research has shown a promising 98,93% in Training Accuracy using IndoBERTtweet.