Article Per Year (5 Year)
p-Index From 2021 - 2026
0.882
P-Index
This Author published in this journals
All Journal Indonesian Journal of Artificial Intelligence and Data Mining JITK (Jurnal Ilmu Pengetahuan dan Komputer) JOURNAL OF SCIENCE AND SOCIAL RESEARCH Jurnal Pengabdian Pada Masyarakat METHABDI
Kelvin Kelvin
Universitas Mikroskil
Religion Agriculture, Biological Sciences & Forestry Computer Science & IT Decision Sciences, Operations Research & Management Economics, Econometrics & Finance Education Immunology & microbiology Languange, Linguistic, Communication & Media Medicine & Pharmacology Physics Public Health Social Sciences

Published : 5 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 5 Documents
Search

 1 
ENHANCING SENTIMENT ANALYSIS ACCURACY WITH BERT AND SILHOUETTE METHOD OPTIMIZATION Kelvin Kelvin; Frans Mikael Sinaga; Wulan Sri Lestari; Sunaryo Winardi; Khairul Hawani Rambe; Ronsen Purba
JITK (Jurnal Ilmu Pengetahuan dan Teknologi Komputer) Vol. 11 No. 1 (2025): JITK Issue August2025
Publisher : LPPM Nusa Mandiri

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33480/jitk.v11i1.6392

Abstract

This research is based on the emergence of ChatGPT technology, which has significant implications in various fields. This research aims to design a model that improves sentiment analysis classification accuracy. The methods applied include the use of the Silhouette Coefficient to determine the best cluster parameters before performing data grouping with the Self-Organizing Map (SOM) method. Additionally, the Bidirectional Encoder Representations from Transformers (BERT) model is utilized to perform precise and convergent sentiment classification. The research methodology encompasses several phases, including data preprocessing through natural language processing techniques. Textual data is converted into vector representations, which are then processed using the Silhouette Coefficient to identify the optimal cluster parameters. These parameters are subsequently applied in the Self-Organizing Map method to cluster data, while the Bidirectional Encoder Representations from Transformers model determines public sentiment, categorized as positive, negative, or neutral. The findings of this study indicate that the best cluster parameter is 9, using a batch size of 64 and a maximum sequence length of 128. The highest accuracy achieved using the confusion matrix is 92.06%. Further tests with varying parameters confirm that the Silhouette Coefficient method significantly enhances the convergence and accuracy of classification outcomes. The conclusion of this research is that integrating the Silhouette Coefficient and Bidirectional Encoder Representations from Transformers is effective in optimizing sentiment analysis on large datasets, achieving both accurate and reliable results.
Kreasi Desain Digital: Pelatihan Figma untuk Siswa SMA Wiyata Dharma Mirza Ilhami; Kristian Telaumbanua; Syanti Irviantina; Irpan Adiputra Pardosi; Kelvin Kelvin
Jurnal Pengabdian Pada Masyarakat METHABDI Vol 5 No 2 (2025): Jurnal Pengabdian Pada Masyarakat METHABDI
Publisher : Universitas Methodist Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.46880/methabdi.Vol5No2.pp343-346

Abstract

In the era of Industry 4.0, the digital divide in high school education, particularly in urban areas like Medan, limits students' access to modern design tools such as Figma for user interface (UI) and user experience (UX) prototyping. This community service activity aims to bridge this gap by providing basic Figma training to 25 Class XI students at SMA Wiyata Dharma. The implementation involved a two-day interactive workshop, covering UI/UX concepts, basic tools (shapes, text, auto-layout), and wireframe/prototype creation for educational apps. Methods included lectures, demonstrations, and hands-on practice, supported by printed modules and free Figma accounts. Evaluation via pre- and post-tests (10 multiple-choice questions) showed a significant improvement from an average score of 45% to 78% (73% increase, p<0.05 via t-test), while partner satisfaction averaged 4.8/5 on a Likert scale. This confirms the hypothesis of at least 25% enhancement in UI/UX understanding, fostering digital creativity and career readiness.
A Smart Architecture for Stunting Prediction: Implementing the SOM–Voting Classifier on Healthcare Big Data Kelvin Kelvin; Sunaryo Winardi; Frans Mikael Sinaga; Hardy Hardy; Erwin Setiawan Panjaitan; Ng Poi Wong; Ferawaty Ferawaty; Justine Lim; Grace Putri Wijaya
Indonesian Journal of Artificial Intelligence and Data Mining Vol. 8 No. 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Childhood stunting is a persistent public health challenge in Indonesia. This study developed a predictive classification model using healthcare data from hospitals in Medan to enable early identification of at-risk children. A novel framework was proposed that integrated an unsupervised Self-Organizing Map (SOM) for feature engineering with a supervised Voting Classifier ensemble, which combined a Support Vector Classifier (SVC), Random Forest (RF), and Gradient Boosting (GB). The proposed framework achieved an accuracy of 100% on the test set, a substantial improvement over the 91.67% accuracy of the baseline Voting Classifier without SOM. While this result highlighted the model's high predictive potential, it must be interpreted cautiously, acknowledging the need for validation on more diverse datasets to ensure generalizability. The findings demonstrated that this hybrid machine learning approach can serve as a powerful decision-support tool, enabling proactive clinical interventions and aiding public health officials in strategically allocating nutritional resources to support Indonesia's national stunting reduction goals.
Big Data Analytics for Predicting Depression Risk in Generation Z: Integrating Self-Organizing Maps and Long Short-Term Memory Joy Nasten Sinaga; Nuraina Nuraina; Frans Mikael Sinaga; Kelvin Kelvin; Nurhayati Nurhayati
Indonesian Journal of Artificial Intelligence and Data Mining Vol. 8 No. 3 (2025): November 2025
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Mental health issues among Generation Z are rising, with depression being one of the most significant challenges. Leveraging the capabilities of big data analytics and artificial intelligence, this study proposes a hybrid method combining Self-Organizing Maps (SOM) and Long Short-Term Memory (LSTM) networks to predict depression risk based on behavioral data. The SOM algorithm is utilized for clustering high-dimensional input data to uncover hidden patterns, while the LSTM network is employed to capture sequential dependencies over time. Data were collected from various digital platforms, processed, and analyzed to train and validate the proposed model. Results show that the SOM-LSTM framework significantly improves the accuracy and reliability of early depression risk detection compared to conventional models. This study contributes a scalable and adaptable model for mental health prediction that can assist in timely interventions for Generation Z
PERBANDINGAN KESESUAIAN PENDAPAT MANUSIA DAN AI (CHATGPT, GEMINI, DEEPSEEK) MENGGUNAKAN PENDEKATAN GROUND TRUTH Kelvin Kelvin; Sunaryo Winardi; Handoko Handoko; Erwin Setiawan Panjaitan; Rivaldi Lubis
JOURNAL OF SCIENCE AND SOCIAL RESEARCH Vol. 9 No. 3 (2026): June 2026
Publisher : Smart Education

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.54314/jssr.v9i3.6784

Abstract

Abstract: The rapid advancement of Large Language Models (LLMs) has significantly improved sentiment analysis capabilities. However, the extent to which these models produce sentiment classifications consistent with human judgment remains an important research question. This study aims to evaluate and compare the agreement of ChatGPT, Gemini, and DeepSeek with human-generated ground truth in sentiment analysis. A total of 1,497 product reviews were collected from the Sephora Products and Skincare Reviews dataset. Three independent annotators labeled each review as positive, neutral, or negative to establish the ground truth. The annotation reliability achieved a Fleiss' Kappa coefficient of 0.7053, indicating substantial agreement and confirming the reliability of the ground truth for evaluation purposes. Subsequently, the three LLMs performed sentiment classification using an Expectation–Role–Action (ERA) prompting strategy. Model performance was assessed using Cohen's Kappa to measure agreement with the ground truth and Spearman's rank correlation to evaluate the consistency of sentiment rankings. The results show that DeepSeek achieved the highest performance, with an average Cohen's Kappa of 0.753 and an average Spearman's correlation coefficient of 0.860, followed by Gemini (κ = 0.662; ρ = 0.834), while ChatGPT demonstrated the lowest agreement (κ = 0.223; ρ = 0.367). These findings indicate that the three LLMs exhibit significantly different levels of agreement with human judgment, with DeepSeek producing sentiment classifications that most closely align with the established ground truth.  Keywords: Large Language Models, Sentiment Analysis, Ground Truth, Cohen's Kappa, Spearman's Rank Correlation.   Abstrak: Perkembangan Large Language Models (LLMs) telah meningkatkan kemampuan analisis sentimen berbasis kecerdasan buatan. Namun, tingkat kesesuaian hasil klasifikasi sentimen yang dihasilkan oleh berbagai LLM terhadap penilaian manusia masih memerlukan evaluasi yang komprehensif. Penelitian ini bertujuan membandingkan tingkat kesesuaian hasil analisis sentimen ChatGPT, Gemini, dan DeepSeek terhadap ground truth yang diperoleh melalui anotasi manusia. Penelitian menggunakan 1.497 ulasan produk dari dataset Sephora Products and Skincare Reviews. Sebanyak tiga anotator independen melakukan pelabelan sentimen ke dalam kategori positif, netral, dan negatif untuk membentuk ground truth. Hasil pengujian reliabilitas menunjukkan nilai Fleiss' Kappa sebesar 0,7053, yang mengindikasikan tingkat kesepakatan Substantial Agreement, sehingga ground truth layak digunakan sebagai acuan evaluasi. Selanjutnya, ketiga model LLM melakukan klasifikasi sentimen menggunakan prompt berbasis Expectation–Role–Action (ERA). Tingkat kesesuaian hasil klasifikasi dievaluasi menggunakan Cohen's Kappa, sedangkan konsistensi hubungan dengan ground truth dianalisis menggunakan korelasi Spearman. Hasil penelitian menunjukkan bahwa DeepSeek memberikan performa terbaik dengan rata-rata Cohen's Kappa sebesar 0,753 dan rata-rata korelasi Spearman sebesar 0,860, diikuti oleh Gemini (κ = 0,662; ρ = 0,834), sedangkan ChatGPT memperoleh tingkat kesesuaian terendah (κ = 0,223; ρ = 0,367). Temuan ini menunjukkan bahwa terdapat perbedaan kemampuan interpretasi sentimen antar model LLM, dengan DeepSeek menghasilkan klasifikasi yang paling mendekati penilaian manusia pada dataset yang digunakan. Kata kunci: Large Language Model, Analisis Sentimen, Ground Truth, Cohen's Kappa, Spearman Correlation.
Co-Authors Erwin Setiawan Panjaitan Erwin Setiawan Panjaitan Ferawaty Ferawaty Frans Mikael Sinaga Grace Putri Wijaya Handoko Handoko Hardy Hardy Irpan Pardosi Joy Nasten Sinaga Justine Lim Khairul Hawani Rambe Kristian Telaumbanua Mirza Ilhami Ng Poi Wong Nuraina nuraina Nurhayati Nurhayati Rivaldi Lubis Ronsen Purba Syanti Irviantina Winardi, Sunaryo Wulan Sri Lestari