DHDedy Hartama

Dedy Hartama

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Dedy Hartama has a total of 141 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Algoritma K-Means Menggunakan Metode Elbow Untuk Mengelompokkan Kinerja Performance Index Berdasarkan Dataset Absensi Pegawai

    by Bunga Sabila, Juwita Permata Sari, Mu’aMmar Nur Kholis, Darman Sahputra Harefa, Dedy Hartama

    Absensi merupakan data kuantitatif yang mencerminkan tingkat kedisiplinan pegawai dan berpotensi digunakan sebagai dasar evaluasi kinerja. Namun, pada banyak instansi, data absensi belum dimanfaatkan secara optimal dalam pengambilan keputusan terkait pengelolaan pegawai. Penelitian ini bertujuan untuk mengelompokkan kinerja pegawai berdasarkan data absensi menggunakan algoritma K-Means Clustering dengan pendekatan Elbow Method. Data yang dianalisis terdiri dari 57 pegawai dengan atribut jumlah keterlambatan, ketidakhadiran, dan absen kosong. Proses penelitian dimulai dari pengumpulan data, pra-pemrosesan, hingga proses klasterisasi. Penentuan jumlah klaster optimal dilakukan dengan Elbow Method menggunakan Python, yang menunjukkan nilai K terbaik adalah empat. Klasterisasi dilakukan menggunakan perangkat lunak RapidMiner, dengan menerapkan alur kerja berbasis visual untuk mengelompokkan pegawai berdasarkan kesamaan pola absensi. Hasil akhir menunjukkan terbentuknya empat klaster dengan karakteristik Key Performance Index (KPI) yang berbeda: sangat baik, baik, cukup, dan kurang. Klaster ini mencerminkan tingkat kedisiplinan pegawai yang dapat digunakan sebagai dasar evaluasi dan pengambilan keputusan manajerial secara objektif. Penelitian ini menunjukkan bahwa penerapan metode klasterisasi berbasis data mining dapat menjadi alat bantu dalam analisis kinerja pegawai di lingkungan organisasi.

  • Open Access

    A Hybrid GRG-Neighborhood Search Model for Dynamic Multi-Depot Vehicle Routing in Disaster Logistics

    by Dedy Hartama, Poningsih Poningsih, Lili Tanti

    A Hybrid GRG-Neighborhood Search Model for Dynamic Multi-Depot Vehicle Routing in Disaster Logistics

  • Open Access

    QUANTUM-ASSISTED FEATURE SELECTION FOR IMPROVING PREDICTION MODEL ACCURACY ON LARGE AND IMBALANCED DATASETS

    by Safii Safii, Mochamad Wahyudi, Dedy Hartama

    One of the biggest obstacles to creating precise machine learning models is choosing representative and pertinent characteristics from big, unbalanced datasets. While too many features raise the risk of overfitting and computational expense, class imbalance frequently results in decreased accuracy and bias. The Simulated Annealing technique is used in this study to tackle a Quadratic Unconstrained Binary Optimization (QUBO) problem that is formulated as a quantum-assisted feature selection method to handle these problems. The technique seeks to reduce inter-feature redundancy and the number of selected features. There are 102,487 samples in the majority class and 11,239 in the minority class, totaling 28 characteristics in the experimental dataset. Nine ideal features were found during the feature selection method (12, 14, 15, 22, 23, 24, 25, 27, and 28). Ten-fold cross-validation was used to assess a Random Forest Classifier that was trained using an 80:20 split. With precision, recall, f1-score, and accuracy all hitting 1.00, the suggested QUBO+SMOTE method demonstrated exceptional performance. Comparatively, QUBO without SMOTE performed worse with accuracy 0.95 and minority-class f1-score of only 0.71, whereas a traditional Recursive Feature Elimination (RFE) approach obtained accuracy 0.97 with minority-class f1-score of 0.94. These findings indicate that QUBO can reduce dimensionality and address class imbalance which requires its integration with SMOTE. This study demonstrates how quantum computing can enhance the effectiveness and efficiency of machine learning, especially for large-scale imbalanced datasets

  • Open Access

    OPTIMIZATION OF EFFICIENTNET-B0 ARCHITECTURE TO IMPROVE THE ACCURACY OF GLAUCOMA DISEASE CLASSIFICATION

    by Imam Akbari, Dedy Hartama, Anjar Wanto

    Glaucoma is a chronic eye disease that can potentially cause permanent blindness if not detected early. This study aims to improve the generalization capability and reliability of glaucoma classification by optimizing the EfficientNetB0 architecture based on a Convolutional Neural Network (CNN). Optimization was carried out by applying double dropout (0.4 and 0.3) and adding a Dense layer with 128 ReLU-activated neurons to reduce overfitting and strengthen non-linear feature representation. The dataset used consists of 1,450 fundus images (899 glaucoma and 551 normal) obtained from IEEE DataPort. Model performance evaluation was performed using accuracy, precision, recall (sensitivity), specificity, F1 score, and Area Under the Curve (AUC) metrics, complemented by confusion matrix analysis to assess overall classification performance. The results showed that the optimized EfficientNetB0 model consistently outperformed the baseline comparison model with the highest accuracy, precision, recall (sensitivity), specificity, F1 score, and AUC values ​​of 95%. Based on the system performance results obtained, the Proposed model can be used as an aid for medical personnel in classifying glaucoma conditions so that they can provide appropriate medical treatment and reduce the risk of permanent blindness due to glaucoma.


  • Open Access

    OPTIMIZING SHUFFLENET WITH GRIDSEARCHCV FOR GEOSPATIAL DISASTER MAPPING IN INDONESIA

    by Abdullah Ahmad, Dedy Hartama, Solikhun Solikhun, Poningsih Poningsih

    Accurate classification of natural disasters is crucial for timely response and effective mitigation. However, conventional approaches often suffer from inefficiency and limited reliability, highlighting the need for automated deep learning solutions. This study proposes an optimized Convolutional Neural Network (CNN) based on the lightweight ShuffleNet architecture, enhanced through GridSearchCV for systematic hyperparameter tuning. Using a geospatial dataset of 3,667 images representing earthquake, flood, and wind-related disasters in Indonesia, the optimized ShuffleNet model achieved a peak accuracy of 99.97%, outperforming baseline CNNs such as MobileNet, GoogleNet, ResNet, DenseNet, and standard ShuffleNet. While these results demonstrate the potential of combining lightweight architectures with automated optimization, the exceptionally high performance also indicates possible risks of overfitting and dataset bias due to limited variability. Therefore, future research should validate this approach using larger, multi-source datasets to ensure robustness and real-world applicability

  • Open Access

    Sentiment Classification in Imbalanced Data: Trade-Offs Between Metrics and Real-World Relevance

    by Indra Swanto Ritonga, Wanayumini, Dedy Hartama

    Sentiment analysis plays a crucial role in assessing public perception, particularly in healthcare services like BPJS Kesehatan, Indonesia’s national health insurance program. However, sentiment classification faces a challenge due to class imbalance, where negative feedback dominates positive responses. This study investigates whether sentiment classification should prioritize traditional evaluation or maintain real-world data representation by preserving the original sentiment distribution. Two feature extraction methods, Term Frequency-Inverse Document Frequency (TF-IDF) and Bag of Words (BoW), were evaluated using Naïve Bayes, Support Vector Machine (SVM), and Logistic Regression with varying maximum feature counts (100–300) to examine the impact of feature dimensionality. Model performance was evaluated using traditional metrics, while sentiment distribution fidelity was assessed by comparing predicted proportions with the dataset. Results show TF-IDF achieves higher precision and recall but fails to capture positive sentiments, leading to a skewed representation of real-world trends, while BoW offers a more balanced distribution with slightly lower accuracy. Paired t-tests and Wilcoxon signed-rank tests confirmed differences in accuracy and recall are significant, but not in precision and sentiment distribution. These findings highlight a trade-off between performance and sentiment diversity, vital in healthcare services and other fields with imbalanced datasets, emphasizing the need to align evaluation metrics with real-world objectives. Future research should investigate advanced models, such as deep learning and transformer-based approaches, to enhance both accuracy and fairness when analyzing imbalanced data.

  • Open Access

    Sentiment Classification on Indodax Using Term Frequency, FastText, and Neural Attention Models

    by Dedy Hartama, Ginanti Riski

    The rapid growth of mobile-based investment platforms such as Indodax has triggered a surge in user-generated reviews that reflect public perception and sentiment. This study aimed to develop and evaluate sentiment classification models that can accurately classify Indonesian user reviews on the Indodax app into negative, neutral, and positive sentiments. A dataset of 11,000 reviews was collected via web scraping from the Google Play Store. Reviews were preprocessed, labeled using a lexicon-based unsupervised method, and balanced using oversampling. Two models were built: a Bidirectional LSTM (BiLSTM) with attention mechanism using FastText embeddings, and a Feedforward Neural Network (FFNN) using a hybrid feature vector combining TF-IDF and FastText. The evaluation was performed using accuracy, classification report, confusion matrix, and PCA visualization. The FFNN model outperformed the BiLSTM-Attention model with an accuracy of 97.07% compared to 96.00%. Both models demonstrated strong performance in classifying three sentiment classes, though the FFNN showed better separation in PCA space and higher macro-average metrics. This study demonstrates the effectiveness of combining statistical and semantic feature representations for sentiment classification in Indonesian text. The proposed approach is particularly valuable for low-resource languages and informal user-generated content.

  • Open Access

    ENHANCED FLOWER IMAGE CLASSIFICATION USING MOBILENETV2 WITH OPTIMIZED PERFORMANCE

    by Ade Ismiaty Ramadhona Ht Barat, Wiwik Sri Astuti, Dedy Hartama, Agus Perdana Windarto, Anjar Wanto

    Flower classification is an essential activity in multiple fields, including healthcare, cosmetics, agriculture, and environmental monitoring. Deep learning has achieved notable success in intricate picture categorization problems, especially through the utilization of lightweight convolutional neural network (CNN) architectures like MobileNet and MobileNetV2. This work assesses and contrasts the efficacy of four prevalent optimizers Adam, RMSProp, SGD, and Nadam on datasets of flower and herbal leaf images. Experiments were performed using a uniform training configuration on a CPU-based system devoid of GPU acceleration, evaluating both model efficacy and computational efficiency. Evaluation criteria including accuracy, precision, recall, F1-score, and loss were utilised, augmented by confusion matrix analysis. The findings indicate that MobileNetV2 regularly surpasses the baseline MobileNet, with RMSProp attaining the highest accuracy (99.52%) and the lowest loss (0.0126) on the herbal dataset. In the flower dataset, RMSProp achieved the highest accuracy of 96.67%. Moreover, MobileNetV2 necessitated increased memory and extended training duration, while delivering superior classification performance overall. These findings underscore the significance of optimizer selection and model architecture in lightweight deep learning applications, especially for deployment on resource-limited devices.


  • Open Access

    OPTIMIZATION OF K-MEANS AND K-MEDOIDS CLUSTERING USING DBI SILHOUETTE ELBOW ON STUDENT DATA

    by Dedy Hartama, Selli Oktaviani

    Clustering methods such as K-Means and K-Medoids are often used to analyze data, including student data, due to their efficiency. However, this method has weaknesses, such as sensitivity to selecting cluster centers (centroids) and cluster results that depend on medoid data. Clustering, an essential technique in data analysis, aims to reveal the natural structure of the data, even in the absence of labeled information. The study, conducted with complete objectivity, compared the performance of two popular clustering methods, K-Means, and K-Medoids, on student data. Three evaluation metrics, namely the Davies-Bouldin Index (DBI), silhouette score, and elbow method, were used to compare clustering and determine the ideal number of clusters for the two algorithms. The data taken in this study are in the form of names, attendance, assignments, formative, midterm exams, final exams, and quality numbers. Based on the existing optimization results, it can be concluded that the K-Means method excels in grouping Student Data. The best results were obtained from the K-Means Algorithm with the Silhouette Coefficient Method with a value of 0.7509 in cluster 2, and the Elbow Method with a value of 1428076.08 in cluster 2, DBI K-Medoids with a value of 0.7413 in cluster 3. So, the best cluster lies in 3 clusters. Keywords: clustering; davies-bouldin indek; elbow method; k-means; k-medoids; silhouette score; Abstrak : Metode clustering seperti K-Means dan K-Medoids sering digunakan untuk menganalisis data, termasuk data siswa, karena efisiensinya. Namun, metode ini memiliki kelemahan, seperti sensitivitas terhadap pemilihan pusat klaster (centroids) dan hasil klaster yang bergantung pada data medoid. Clustering, sebuah teknik penting dalam analisis data, bertujuan untuk mengungkapkan struktur alami dari data, bahkan tanpa adanya informasi berlabel. Penelitian ini, yang dilakukan dengan objektivitas penuh, membandingkan kinerja dua metode clustering populer, yaitu K-Means dan K-Medoids, pada data mahasiswa. Tiga metrik evaluasi, yaitu Davies-Bouldin Index (D.B.I.), silhouette score, dan metode elbow, digunakan untuk membandingkan clustering dan menentukan jumlah cluster yang ideal untuk kedua algoritma tersebut. data yang diambil dalam penelitian ini berupa nama, kehadiran, tugas, formatif, ujian tengah semester, ujian akhir semester, angka mutu. Berdasarkan hasil optimasi yang ada, dapat disimpulkan bahwasannya metode K-Means unggul dalam pengelompokkan Data Mahasiswa. Sehingga di peroleh hasil terbaik dari Algoritma K-Means dengan Metode Silhouette Coefficient dengan nilai 0,7509 di cluster 2, dan Elbow Method dengan nilai 1428076,08 di cluster 2, DBI K-Medoids dengan nilai 0,7413 di cluster 3. Sehingga cluster terbaik terletak pada 3 cluster. Kata kunci: klasterisasi; davies-bouldin indek; elbow method; k-means; k-medoids; silhouette score;


  • Open Access

    OPTIMIZING THE KNN ALGORITHM FOR CLASSIFYING CHRONIC KIDNEY DISEASE USING GRIDSEARCHCV

    by Muhammad Rahmansyah Siregar, Dedy Hartama, Solikhun Solikhun

    Chronic Kidney Disease (CKD) is a progressive condition that impairs kidney function and cannot be cured. Early detection is crucial for effective management and therapy. However, diagnosing CKD is challenging as patients often have comorbidities such as diabetes, hypertension, or heart disease, which complicate diagnosis and treatment. Accurate classification methods are essential for early detection. K-Nearest Neighbor (KNN) is a classification algorithm that groups data based on feature similarity. K-NN is an algorithm that is resistant to outliers, easy to implement, and highly adaptable. It only requires distance calculations between data points and does not involve complex parameters. However, its performance depends on hyperparameters such as the number of neighbors (k), weighting, and distance metric. Incorrect hyperparameter selection can lead to overfitting, underfitting, or reduced accuracy. To address these issues, GridSearchCV is used to optimize KNN by systematically selecting the best hyperparameters, ensuring improved accuracy and reduced overfitting. This optimization enhances the model’s reliability in early CKD detection compared to other methods. This study aims to determine the optimal KNN parameters for CKD classification using GridSearchCV. The results show 8.05% accuracy improvement and reduction in overfitting, with the prediction gap between training and testing decreasing from 6% to only 1.15%. These enhancements contribute to more reliable CKD diagnosis, enabling accurate early detection and better clinical decision-making.