ejournal.nusamandiri.ac.id (STMIK Nusa Mandiri)
Not a member yet
1504 research outputs found
Sort by
OPTIMIZATION CVRP WITH MACHINE LEARNING FOR IMPROVED CLASSIFICATION OF IMBALANCED DATA FOOD DISTRIBUTION
Get PDFThe classification of imbalanced data in food delivery distribution is an important issue that needs to be considered to ensure fairness and efficiency in the food distribution system. This research answers these problems by improving the accuracy of the classification of delivery locations that have imbalanced demand data, so that high priority areas are not neglected. Generating more efficient and cost-effective distribution routes, taking into account vehicle capacity and delivery urgency. Reducing delivery time and potential food waste due to delays or non-optimal route allocation. This study addresses the problem of improving the accuracy of delivery location classification that has imbalanced demand data, so that high priority areas are not neglected. Generate more efficient and cost-effective distribution routes, taking into account vehicle capacity and delivery urgency. Reduce delivery time and potential food wastage due to delays or non-optimal route allocation. This study uses the research stages of data collecting, data preprocessing, and implementation of K-Means and K-NN methods. The results of CVRP testing with K-Means show the value of cluster 7 acc=80, precc=85, recall=84. cluster 9 acc=85, precc=90, recall=91. cluster 11 acc=88, precc=93, recall=94. While the results of CVRP testing with K-NN show the value of K 7 acc=89, precc=88, recall=85. value of K 9 acc=87, precc=90, recall=91. value of K 11 acc=95, precc=97, recall=94. The optimization results show that this approach not only improves operational efficiency but also increases the accuracy of food delivery, which will affect the availability of traditional markets
OPTIMIZATION OF THE INCEPTIONV3 ARCHITECTURE FOR POTATO LEAF DISEASE CLASSIFICATION
Get PDFPotato leaf diseases can cause significant yield losses, making early detection crucial to prevent major damages. This study aims to optimize the Inception V3 architecture in a Convolutional Neural Network (CNN) for potato leaf disease classification by applying Fine Tuning Pre-Trained. This method leverages weights from a pre-trained model on a large-scale dataset, enhancing accuracy while reducing the risk of overfitting. The training process involves adjusting several final layers of Inception V3 to better adapt to specific features of potato leaf diseases. The results show that this approach improves classification performance, achieving an accuracy of 97.78%, precision of 98%, recall of 98%, and an F1-score of 98%. With better computational efficiency compared to previous architectures, this model is expected to be widely applicable in plant disease detection systems, particularly for farmers or institutions with limited resources
A LIGHTWEIGHT AND PRACTICAL PIPELINE FOR CROSS-PROJECT DEFECT PREDICTION USING METRIC-BASED LEARNING
Get PDFCross-Project Defect Prediction (CPDP) addresses the scarcity of defect data in new software projects by transferring knowledge from existing ones. However, domain shift between projects remains a major challenge. This study introduces a lightweight and practical CPDP pipeline based on traditional metric features, integrating domain adaptation (CORAL, TCA, TCA+), feature selection, and resampling techniques. Through 120 configurations evaluated on multiple PROMISE datasets, we found that combining TCA or TCA+ with Synthetic Minority Over-sampling Technique combined with Edited Nearest Neighbors (SMOTEENN) consistently improved F1-Score and Recall on imbalanced datasets. LightGBM demonstrated the most stable performance across projects, while Logistic Regression yielded the highest MCC in specific cases. Principal Component Analysis (PCA) visualizations supported the effectiveness of domain alignment. The proposed pipeline offers a reproducible, cost-efficient alternative to deep learning models and provides actionable insights for defect prediction in resource-constrained environments
THE IMPACT OF COLOR AND CONTRAST ENHANCEMENT FOR DIAGNOSING GASTROINTESTINAL DISEASES BASED DEEP LEARNING
Get PDFEndoscopy is a crucial tool for diagnosing digestive tract diseases—colon cancer and polyps using a camera with LED lighting, but often results in low-quality images with poor contrast and luminance. This study evaluates the performance of two contrast-based image quality enhancement—Contrast Limited Adaptive Histogram Equalization (CLAHE) and Improved Adaptive Gamma Correction with Weighting Distribution (IAGCWD)—along with various color space transformations (RGB, HSV, YCbCr, CIELAB, Grayscale) in deep learning-based digestive tract diseases detection system. The detection system using EfficientNetV2S model and Quadratic Weighted Kappa (QWK) loss function to obtain the balance of prediction results for each class. The experiment shows that CLAHE is able to achieve 79% accuracy which is superior in clarifying important information in endoscopy images. CLAHE performs well due to its ability to reduce noise and enhance contrast. The classification model with HSV and CLAHE on KVASIR is able to recognize all classes well. RGB, HSV, and YCbCr color spaces have stable performance in most tests. This study contributes insights for enhancing endoscopic image quality to support both computer-aided and clinical diagnosis
PERFORMANCE COMPARISON OF DEEP CNN ARCHITECTURES FOR LUNG AREA SEGMENTATION IN CHEST IMAGING
Get PDFLung area segmentation is a critical preprocessing step in computer-aided diagnosis systems for respiratory diseases such as lung cancer and pneumonia. Accurate segmentation enhances the detection and monitoring of pathological conditions but manual delineation is time-consuming and subject to variability. This research aims to identify the most effective convolutional neural network (CNN) architecture for automated lung segmentation by comparing three models: U-Net, DeepLab, and a proposed hybrid model combining U-Net with ResUNet_Light. The models were trained and evaluated using a publicly available chest CT dataset under identical experimental settings, including preprocessing steps, training parameters, and standard evaluation metrics: Dice Similarity Coefficient (DSC), Intersection over Union (IoU), Precision, and Recall. Results show that the proposed U-Net + ResUNet_Light model achieves the best performance across all metrics (DSC: 0.6767, IoU: 0.5652, Precision: 0.8480, Recall: 0.7920), outperforming both U-Net and DeepLab. These improvements are attributed to the integration of residual blocks, which enhance feature propagation and gradient flow, enabling better generalization and segmentation accuracy, especially along complex lung boundaries. In contrast, while DeepLab performs well in capturing contextual information, its higher complexity may hinder real-time applicability. U-Net, though efficient, showed limitations in accurately segmenting irregular regions. The findings demonstrate the potential of the proposed model for clinical deployment, where both accuracy and efficiency are critical. This study contributes to the development of more robust deep learning-based segmentation methods and highlights the importance of architectural enhancements in CNN design for medical image analysis
OPTIMIZATION OF STUNTING INFANT DATA CLUSTERING WITH K-MEANS++ ALGORITHM USING DBI EVALUATION
Get PDFStunting in infants is a serious health issue, particularly in developing countries like Indonesia. This study aims to optimize the clustering of stunting data in infants using the K-Means++ algorithm, evaluated with the Davies-Bouldin Index (DBI) to determine the optimal number of clusters. The stunting data includes variables such as age, gender, weight, and height. The analysis results indicate that the optimal number of clusters is 5, with a DBI value of 0.837986204, confirming the quality of the clustering. This conclusion demonstrates that the combination of these evaluation methods produces effective clustering and provides significant insights into identifying groups of infants with varying stunting risk levels. These findings can serve as a basis for more targeted health interventions in addressing stuntin
PENGEMBANGAN CHATBOT TELEGRAM FAQ LAYANAN ICT MENGGUNAKAN ALGORITMA RANDOM FOREST DAN METODE WORD2VEC
Get PDFIn today's digital era, chatbots have become an essential tool for businesses to improve interaction with customers. A responsive and efficient chatbot can help customer service agents be happier, improve customer satisfaction, and resolve issues faster. The study aims to create a Telegram-based chatbot that uses the Random Forest algorithm and the Word2Vec method to answer questions about ICT services. The development was carried out by collecting a dataset of questions and answers from FAQs (Frequently Asked Questions) of ICT services. Then, the Random Forest algorithm is used to classify the questions. In addition, the Word2Vec method is used to create vector representations of words in questions and answers. This improves the chatbot's ability to understand complex questions. The test results show that the chatbot gets an accuracy of 91.28%, a precision of 93.56%, a recall of 91.28% and an F1-Score of 91.42% and can provide relevant and accurate answers to user questions. Therefore, the development of this chatbot using the Random Forest algorithm and the Word2Vec method can be an effective solution to improve customer service in the field of ICT service
RANCANG BANGUN APLIKASI PENGELOLAAN PPH 21 PADA CV.ECS CONSULTING SERVICES DENGAN PENDEKATAN RAD
Get PDFIncome Tax (PPh) 21 often poses a challenge for companies and employees in managing tax payments efficiently and accurately. CV. ECS Consulting Service, which currently uses Microsoft Excel for PPh 21 calculations, faces risks of errors and time-consuming manual processes. This research aims to develop a web-based PPh 21 calculation application using the Rapid Application Development (RAD) method tailored to the company’s needs. Data collection methods include direct observation and literature study, while the application development method adopts RAD, which is iterative and responsive to changing requirements. The scope of the research includes user interface design, development of calculation algorithms, employee data processing, and ensuring data security. The application was tested using Black Box testing to ensure all features function properly, User Acceptance Testing (UAT) to assess whether it meets user needs, and performance testing to evaluate the website’s speed and stability. Black Box testing was conducted on six cases, and UAT was carried out directly with users. The results showed that the application passed all tests and met the required functionalities. Performance testing also indicated that the system is fairly stable, although further improvement is needed for long-term use or during high traffic
IMPROVING HANDWRITTEN DIGIT RECOGNITION USING CYCLEGAN-AUGMENTED DATA WITH CNN–BILSTM HYBRID MODEL
Get PDFHandwritten digit recognition presents persistent challenges in computer vision due to the high variability in human handwriting styles, which necessitates robust generalization in classification models. This study proposes an advanced data augmentation strategy using Cycle-Consistent Generative Adversarial Networks (CycleGAN) to improve recognition accuracy on the MNIST dataset. Two architectures are evaluated: a standard Convolutional Neural Network (CNN) and a hybrid model combining CNN for spatial feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) for sequential pattern modeling. The CycleGAN-based augmentation generates realistic synthetic images that enrich the training data distribution. Experimental results demonstrate that both models benefit from the augmentation, with the CNN-BiLSTM model achieving the highest accuracy of 99.22%, outperforming the CNN model’s 99.01%. The study’s novelty lies in the integration of CycleGAN-generated data with a CNN–BiLSTM architecture, which has been rarely explored in previous works. These findings contribute to the development of more generalized and accurate deep learning models for handwritten digit classification and similar pattern recognition tasks
REFORMULATION OF MULTI-ATTRIBUTE UTILITY THEORY NORMALIZATION TO HANDLE ASYMMETRIC DATA IN MADM
Get PDFMulti-Attribute Utility Theory (MAUT) is a widely used multi-attribute decision-making (MADM) method due to its ability to integrate multiple criteria into a single utility value. However, conventional MAUT faces limitations when handling asymmetric data, where standard normalization processes often lead to value distortion and less representative rankings. This study aims to reformulate the normalization function in MAUT to improve adaptability to non-symmetric data distributions and to enhance ranking validity in decision-making. A modification approach called MAUT-A was developed by applying an adaptive normalization mechanism capable of accommodating extreme distributions and outliers by adding Z-score normalization. The performance of MAUT-A was evaluated by comparing the correlation of its ranking results with reference rankings, and the outcomes were benchmarked against conventional MAUT. The experimental findings indicate that conventional MAUT achieved a correlation value of 0.9688 with the reference ranking, while the proposed MAUT-A method achieved a higher correlation of 0.9792. This improvement represents that MAUT-A has better suitability, stability, and reliability in managing asymmetric data. The study contributes by offering a reformulated MAUT framework through adaptive normalization, providing more accurate, stable, and fair ranking outcomes. This approach enhances the validity of MADM applications, particularly in contexts involving asymmetric data distribution