Article Per Year (5 Year)
p-Index From 2021 - 2026
1.197
P-Index
This Author published in this journals
All Journal Media Statistika JOURNAL OF APPLIED INFORMATICS AND COMPUTING Jurnal Informatika: Jurnal Pengembangan IT Parameter: Jurnal Matematika, Statistika dan Terapannya
Suhaeni, Cici
Unknown Affiliation
Computer Science & IT Decision Sciences, Operations Research & Management Mathematics

Published : 7 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 7 Documents
Search

 1 
Pemodelan Topik pada Komentar YouTube Arra: Komparasi LDA dan K-Means Menggunakan Fitur Leksikal dan Semantik Nuradilla, Siti; Kamila, Sabrina Adnin; Zahra, Latifah; Suhaeni, Cici; Sartono, Bagus
Jurnal Informatika: Jurnal Pengembangan IT Vol 10, No 3 (2025)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v10i3.8763

Abstract

YouTube has become a platform for sharing content, including positive material and stereotypes that often trigger debates. One noteworthy phenomenon is the video of Arra, a toddler known for her remarkable communication skills. This uniqueness has drawn significant attention and sparked debates about the mismatch between her age and cognitive development. The diverse comments on Arra’s videos reflect sharply differing perspectives among netizens, making manual analysis highly challenging. Therefore, it is important to examine the topics discussed by netizens to understand the dominant issues emerging in these discussions. Through this approach, the public can gain insights, and parents may receive valuable input regarding child-rearing practices. The main objective of this study is to explore the effectiveness of the two methods and their combinations of text representations in identifying key topics within comments by comparing the coherence performance of the models. This research applies topic modeling to analyze comments using two primary approaches: Latent Dirichlet Allocation (LDA) and K-Means clustering. The study involves data collection through comment crawling, followed by text preprocessing and text representation using TF-IDF and GloVe embeddings. LDA and K-Means are then used to identify dominant topics appearing in the comments. The results show that LDA with TF-IDF achieved the highest coherence score of 0.662, although the resulting topics were still difficult to interpret due to overlap. Meanwhile, K-Means with GloVe 100D yielded a slightly lower coherence score of 0.6538 but outperformed in terms of interpretability. Therefore, K-Means with GloVe 100D is considered a more balanced approach in terms of both coherence and topic readability.
Comparison of LASSO, Ridge, and Elastic Net Regularization with Balanced Bagging Classifier Nisrina Az-Zahra, Putri; Sadik, Kusman; Suhaeni, Cici; Mohamad Soleh, Agus
Parameter: Jurnal Matematika, Statistika dan Terapannya Vol 4 No 2 (2025): Parameter: Jurnal Matematika, Statistika dan Terapannya
Publisher : Jurusan Matematika FMIPA Universitas Pattimura

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30598/parameterv4i2pp287-296

Abstract

Predicting Drug-Induced Autoimmunity (DIA) is crucial in pharmaceutical safety assessment, as early identification of compounds with autoimmune risk can prevent adverse drug reactions and improve patient outcomes. Classification analysis often faces challenges when the number of predictor variables exceeds the number of observations or when high correlations among predictors lead to multicollinearity and overfitting. Regularization methods, such as Ridge Regression, Least Absolute Shrinkage and Selection Operator (LASSO), and Elastic-Net, help stabilize parameter estimation and improve model interpretability. This study focuses on building a binary classification model to predict the risk of DIA using 196 molecular descriptors derived from chemical compound structures. To address class imbalance in the response variable, the Balanced Bagging Classifier (BBC) is combined with regularized logistic regression models. Elastic Net + BBC outperforms other models with the highest accuracy (0.825), followed closely by LASSO + BBC and Ridge + BBC (both 0.816). This integration not only improves classification accuracy but also enhances generalization and the reliable detection of minority class instances, supporting the early identification of autoimmune risks in drug discovery.
EVALUATING RANDOM FOREST AND XGBOOST FOR BANK CUSTOMER CHURN PREDICTION ON IMBALANCED DATA USING SMOTE AND SMOTE-ENN Andespa, Reyuli; Sadik, Kusman; Suhaeni, Cici; Soleh, Agus M
MEDIA STATISTIKA Vol 18, No 1 (2025): Media Statistika
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/medstat.18.1.25-36

Abstract

The banking industry faces significant challenges in retaining customers, as churn can critically affect both revenue and reputation. This study introduces a robust churn prediction framework by comparing the performance of XGBoost and Random Forest algorithms under imbalanced data conditions. The novelty of this research lies in integrating the SMOTE and SMOTE-ENN techniques with machine learning algorithms to enhance model performance and reliability on highly imbalanced datasets. Unlike conventional approaches that rely solely on oversampling or undersampling, this study demonstrates that the hybrid combination of XGBoost and SMOTE provides superior predictive accuracy, stability, and efficiency. Hyperparameter optimization using GridSearchCV was conducted to identify the most effective parameter configurations for both algorithms. Model performance was evaluated using the F1-Score and Area Under the Curve (AUC). The results indicate that XGBoost with SMOTE achieved the best performance, with an F1-Score of 0.8730 and an AUC of 0.9828, showing an optimal balance between precision and recall. Feature importance analysis identified Months_Inactive_12_mon, Total_Trans_Amt, and Total_Relationship_Count as the most influential predictors. Overall, this approach outperforms traditional resampling and modeling techniques, providing practical insights for data-driven customer retention strategies in the banking industry.
Evaluating Fasttext and Glove Embeddings for Sentiment Analysis of AI-Generated Ghibli-Style Images Sentana Putra, I Gusti Ngurah; Yusran, Muhammad; Sari, Jefita Resti; Suhaeni, Cici; Sartono, Bagus; Dito, Gerry Alfa
Journal of Applied Informatics and Computing Vol. 9 No. 5 (2025): October 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i5.10600

Abstract

The development of text-to-image generation technology based on artificial intelligence has triggered mixed public reactions, especially when applied to iconic visual styles such as Studio Ghibli. This research aims to evaluate public sentiment towards the phenomenon of Ghibli-style AI images by comparing two static word embedding methods, namely FastText and GloVe, on three classification algorithms: Logistic Regression, Random Forest, and Convolutional Neural Network (CNN). Data in the form of Indonesian tweets were collected from Twitter using hashtags such as #ghibli, #ghiblistyle, and #hayaomiyazaki during the period 25 March to 25 April 2025. Each tweet was manually labelled with positive or negative sentiment, then preprocessed and represented using pre-trained FastText and GloVe embeddings. Evaluation was conducted using accuracy, precision, recall, and F1-score metrics, both macro and weighted. Results showed that FastText consistently performed the best on most models, especially in terms of precision and overall accuracy, thanks to its ability to handle sub-word information and spelling variations in social media texts. The combination of CNN with FastText yielded the highest performance with a macro F1-score of 76.56% and accuracy of 84.69%. However, GloVe still showed competitive performance in recall on the Logistic Regression model, making it relevant for contexts that prioritise sentiment detection coverage. This study emphasizes the importance of selecting embeddings and models that are appropriate to the characteristics of the data and the purpose of the analysis in informal social media-based sentiment classification.
The Impact of the L1/L2 Ratio on Selection Stability and Solution Sparsity along the Elastic Net Regularization Path in High-Dimensional Genomic Data Fahira, Fani; Sadik, Kusman; Suhaeni, Cici; M Soleh, Agus
Journal of Applied Informatics and Computing Vol. 10 No. 1 (2026): February 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i1.12059

Abstract

High-dimensional genomic datasets (p>n) pose persistent challenges for predictive modeling and biomarker-oriented feature selection due to multicollinearity and instability of selected feature sets under resampling. Although Elastic Net is widely used to address correlated predictors via combined L1/L2 regularization, the practical role of the L1/L2 mixing ratio (α) is often treated as a secondary tuning choice driven primarily by predictive accuracy. This study investigates how varying α shapes the trade-off among selection stability, solution sparsity, and predictive performance along the Elastic Net regularization path. Experiments were conducted using the publicly available METABRIC breast cancer cohort (n = 1,964) with 21,113 gene expression features and a binary overall survival status outcome. Logistic regression with Elastic Net penalty was fitted across a grid of α values, with the regularization strength (λ) selected by cross-validation. Feature selection stability was evaluated under repeated resampling using the Jaccard index, Dice coefficient, and Adjusted Rand Index (ARI), while sparsity was summarized by the average number of non-zero coefficients; predictive performance was assessed using AUC, accuracy, and F1-score. Results show a monotonic decline in stability as α increases: α = 0.2 yields the highest stability (Jaccard 0.324, Dice 0.487, ARI 0.434), whereas LASSO (α = 1.0) produces the lowest stability (Jaccard 0.278, Dice 0.431, ARI 0.400). In contrast, predictive performance varies only marginally across α (AUC 0.696–0.704; accuracy 0.666–0.671; F1-score 0.738–0.742), while sparsity changes substantially (average selected features 110–204). Coefficient path analyses further illustrate abrupt shrinkage under LASSO versus smoother, group-preserving shrinkage under Elastic Net, consistent with improved reproducibility under lower-to-moderate α. Frequency-of-selection analysis highlights genes repeatedly selected across resampling, supporting interpretability of stable configurations without claiming causal biomarker validity. Overall, the findings demonstrate that α is a substantive modeling choice that materially affects stability and sparsity even when accuracy is similar, motivating stability-aware tuning for high-dimensional genomic prediction and reproducible feature discovery.
From Sparse Features to Transformers: A Statistical Evaluation of TF-IDF, FastText, and IndoBERT for Sentiment Classification of Indonesian Travel App Reviews Tangdilomban, Claudian Tikulimbong; Ramdhan, Syaifullah Yusuf; Rizal, Muhammad; Suhaeni, Cici; Sartono, Bagus
Journal of Applied Informatics and Computing Vol. 10 No. 3 (2026): June 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i3.12610

Abstract

This study compares three text representation techniques, namely TF-IDF, FastText, and IndoBERT, in the sentiment classification task of Indonesian-language user reviews of travel applications. The dataset consists of 4.000 reviews from Traveloka and Tiket.com, collected through Google Play Store scraping and manually annotated with sentiment labels. Each representation technique was combined with three classification algorithms, namely Support Vector Machine, Logistic Regression, and Random Forest, resulting in nine experimental configurations. The evaluation was conducted using stratified 5-fold cross-validation with macro F1-score as the primary metric, supported by hyperparameter tuning using GridSearchCV, paired t-test statistical analysis, and Cohen’s d effect size measurement. The evaluation results indicate that IndoBERT generally achieved the best performance compared to TF-IDF and FastText. The best configuration was obtained by IndoBERT with Logistic Regression, achieving an F1-score of 0.9261 after tuning. The statistical test showed that the performance differences among text representations were statistically significant, with large effect sizes in the comparison between IndoBERT and TF-IDF (d = −1.36) and between IndoBERT and FastText (d = −1.10). Nevertheless, TF-IDF combined with Logistic Regression and SVM remained competitive, achieving an F1-score of approximately 0.892 after tuning, making it a lightweight and interpretable alternative. This study concludes that the quality of text representation has a more dominant influence on sentiment classification performance than the complexity of the classification algorithm.
Preprocessing Effects on Word2Vec and IndoRoBERTa for Shopee Sentiment Classification Avitadia Prichanti, Lilik; Rizky Pangestutik, Kinanti; Suhaeni, Cici; Sartono, Bagus
Journal of Applied Informatics and Computing Vol. 10 No. 5 (2026): October 2026
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v10i5.13014

Abstract

This study evaluates the effects of heavy and minimal preprocessing on static and contextual text representations for sentiment classification of Indonesian Shopee reviews. The dataset comprised 29,618 Google Play Store reviews collected in January 2026 and weakly labeled from star ratings into negative, neutral, and positive classes. The data were divided into training, validation, and test sets using an 80:10:10 stratified split. Heavy preprocessing included case folding, text cleaning, slang normalization, stopword removal, and stemming, whereas minimal preprocessing applied case folding only. Four experimental configurations combined the two preprocessing strategies with Word2Vec–Logistic Regression and fine-tuned IndoRoBERTa. Model performance was evaluated using accuracy, weighted precision, weighted recall, weighted F1-score, macro F1-score, and confusion matrices. Heavy preprocessing produced a marginal weighted F1-score improvement for Word2Vec–Logistic Regression, from 86.0% to 86.2%. Minimal preprocessing improved IndoRoBERTa’s weighted F1-score from 88.4% to 89.7% and achieved the best accuracy of 90.7% and macro F1-score of 65.8%. The difference between weighted and macro F1-scores indicates that the neutral minority class remained difficult to identify. The findings show that preprocessing effectiveness depends on model architecture. Minimal preprocessing better preserves contextual information for IndoRoBERTa, whereas aggressive preprocessing provides only a marginal benefit for the static baseline.
Co-Authors Agus Mohamad Soleh Andespa, Reyuli Avitadia Prichanti, Lilik Bagus Sartono Dito, Gerry Alfa Fahira, Fani Kamila, Sabrina Adnin Kusman Sadik M Soleh, Agus MUHAMMAD RIZAL Muhammad Yusran Nisrina Az-Zahra, Putri NURADILLA, SITI Ramdhan, Syaifullah Yusuf Rizky Pangestutik, Kinanti Sari, Jefita Resti Sentana Putra, I Gusti Ngurah Tangdilomban, Claudian Tikulimbong Zahra, Latifah