Journal of Informatics And Telecommunication Engineering
Not a member yet
373 research outputs found
Sort by
The Implementation of Random Forest to Predict Sales a Case Study at Chatime Binjai Supermall
Full text linkIn an increasingly competitive business environment, retail industries like Chatime Binjai Supermall must quickly adapt. Changes in consumer trends, preferences, and technological advancements significantly impact business strategies. To stay competitive, Chatime Binjai Supermall needs to optimize sales, marketing, and inventory management through accurate data analysis and prediction. Random Forest, a powerful machine learning algorithm, is used to process historical data and more accurately predict sales. This study evaluates the performance of Random Forest in predicting daily, weekly, and monthly sales. The analysis shows that products like "Jasmine Green Tea (L)" have the highest daily demand, "PEARL (L)" leads weekly sales, and there is an increase in demand for specific products monthly, such as "CT RAINBOW JELLY (L)." The implementation of the Random Forest algorithm at Chatime Binjai Supermall demonstrates significant potential in enhancing sales efficiency and data-driven decision-making, helping the company remain relevant and competitive amidst market changes
Improving the Accuracy of Coffee Leaf Disease Detection Using Squeezenet and Simam
Full text linkEarly detection of coffee leaf diseases such as leaf rust and Phoma is essential due to its direct impact on crop productivity and quality. Recent studies have shown that lightweight CNN architectures like SqueezeNet are effective for deployment on resource-constrained devices, though they still face limitations in classification accuracy for complex disease types. This study aims to improve the accuracy of coffee leaf disease classification by integrating the SqueezeNet architecture with the SimAM attention module, which enhances feature representation without significantly increasing model complexity. A quantitative experimental approach was used, employing an open-source dataset of coffee leaf images that was augmented and categorized into three classes: healthy leaves, leaf rust, and Phoma. The models were evaluated using accuracy, precision, recall, and F1-score metrics. Results show that integrating SimAM into SqueezeNet increased the model’s accuracy from 81% to 84%. The most significant improvements were observed in the leaf rust and Phoma classes, with F1-scores rising from 0.70 to 0.79 and from 0.73 to 0.76, respectively. Additionally, the AUC score improved to 0.91. These results demonstrate that SimAM integration effectively enhances classification performance, though challenges remain in distinguishing classes with visually similar features. Further research is recommended to implement more aggressive data augmentation and regularization techniques to improve model generalization
Efficient Real and Fake Face detection Using ResNet18
Full text linkThis study aims to develop a classification model for distinguishing between real and fake facial images using a lightweight Convolutional Neural Network architecture, specifically ResNet18. The research addresses the growing misuse of synthetic facial images in biometric security systems and identity verification processes. A combined dataset was used, consisting of secondary data from the 140K Real and Fake Faces dataset on Kaggle and primary images captured via a local camera. Preprocessing steps included resizing all images to 128×128 pixels, horizontal flipping, and normalization. The model was trained for five epochs using the FastAI framework with the one-cycle learning rate strategy. The experimental results show that the ResNet18 model achieved a test accuracy of 92.1%, with balanced precision, recall, and F1-score across both classes. Evaluation metrics were supported by a classification report and confusion matrix. The model contains 11.7 million parameters and completed training in approximately 9 minutes and 42 seconds, indicating its computational efficiency on a T4 GPU environment. While the study referenced deeper architectures such as ResNet34 and ResNet50 for context, no direct comparative experiments were conducted. Therefore, conclusions regarding relative performance are limited to the reported metrics of ResNet18 alone. The findings support the feasibility of deploying ResNet18-based models for real-time facial image classification in resource-constrained environments. Future research is encouraged to explore architecture comparisons, more advanced augmentation techniques, and evaluation using video-based inputs for improved generalizatio
Comparative Analysis Using Xception and MobileNetV2 Deep Learning Models for Brain Tumor Detection in MRI Images
Full text linkThis study presents a comparative analysis of two deep learning models, Xception and MobileNetV2, for brain tumor detection using MRI images. The selection of these models is based on their respective advantages. Xception is known for its ability to handle large and complex datasets due to its deep architecture and the use of depthwise separable convolutions. It also features a deep structure capable of extracting complex features from high-resolution images, making it well-suited for detailed image recognition tasks. In contrast, MobileNetV2 is designed to be lighter and more computationally efficient, making it ideal for deployment on mobile devices or in resource-constrained environments without significantly compromising performance. These characteristics make both models highly relevant for medical image analysis, particularly in brain tumor detection, which demands both accuracy and efficiency.This study uses a public dataset that has been preprocessed through augmentation and normalization. Both models were trained and evaluated using accuracy, loss, and confusion matrix metrics. The results show that MobileNetV2 achieved higher accuracy (97.8%) compared to Xception (94.9%) with a lower error rate. For precision, recall, and F1-score metrics, the results were identical up to four decimal places, further supporting that MobileNetV2 is more suitable for brain tumor detection in resource-limited settings. Based on the findings, MobileNetV2 demonstrates superior performance compared to Xception, making it the favorable choice
Port Risk Mitigation with FMEA Method on Port Operational Information System at PT. Pelindo (Persero) Sibolga Branch: Case Study at Port of Sibolga
No full textPort operations face challenges in the form of potential risks such as delays in data recording, data inconsistencies between units, and lack of system integration that can hinder logistics distribution. This study identified 20 potential operational risks using the Failure Mode and Effect Analysis (FMEA) method to help map mitigation priorities through the calculation of the Risk Priority Number (RPN). The results of the risk mapping were used as a basis for designing the functional requirements of a web-based port operational information system. The system was developed using PHP, Laravel, and MySQL to support structured recording of loading and unloading activities, ship scheduling, and logistics monitoring. Although the RPN values were used to understand risk priorities, they did not directly determine the system features. Instead, the risk analysis served to provide an overall understanding for designing a system that better matches operational needs. The validation of system benefits at this stage remains conceptual, and future implementation is needed to test its effectiveness in actual port operations
Analysis Role of Digital Marketing and Self-Image Improving High School Students' Self-Presentation in Batam Using Instagram
Full text linkWith the rapid development of digital technology, Instagram serves as a critical platform for shaping self-image and promoting self-presentation among students. This study explores the influence of digital marketing and self-image on the self-presentation of high school students in Batam through Instagram, with Digital marketing and self-image serves as the independent variable and self-presentation as the dependant variable. Using a quantitative research method, data were collected from 353 respondents across 16 private high schools in Batam. The study finds that both digital marketing and self-image significantly affect students' self-presentation, with digital marketing enabling creative self-promotion and self-image enhancing confidence and personality expression. Key Instagram features like stories, highlights, reels, and posts are instrumental in fostering positive self-image and social interaction. Statistical analysis revealed that 22.8% of self-presentation is influenced by the studied variables, highlighting the role of external factors. This research underscores Instagram's role in inspiring fashion trends, promoting products, and cultivating a polished appearance, contributing to students' confidence and social perception. These findings have implications for understanding digital media's impact on youth behaviour and social identit
Performance Evaluation of CNN-LSTM and CNN-FNN Combinations for Pneumonia Classification Using Chest X-ray Images
Full text linkPneumonia is one of the deadliest infectious diseases worldwide, particularly affecting children under five years old and the elderly, with a significant mortality rate annually. This disease is caused by bacterial, viral, or fungal infections, leading to inflammation in the air sacs (alveoli) of the lungs, which disrupts respiratory function. A major challenge in diagnosing pneumonia lies in the reliance on radiological expertise to interpret chest X-ray images, a process that is time-consuming and prone to errors in interpretation. This study aims to compare the performance of deep learning models, specifically the combination of Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) and CNN with Feedforward Neural Networks (FNN), in classifying pneumonia based on chest X-ray images. The results indicate that the CNN & LSTM model achieved an accuracy of 96.59%, a loss of 9.95%, precision of 96%, recall of 95%, and F1-score of 96%, slightly outperforming the CNN & FNN model, which achieved an accuracy of 96.13%, a loss of 12.16%, precision of 96%, recall of 94%, and F1-score of 95%. The advantage of CNN & LSTM lies in its ability to capture sequential patterns through LSTM, making it more effective in detecting positive pneumonia cases. In conclusion, the CNN & LSTM model outperforms the CNN & FNN model in accuracy, recall, and F1-score, making it a more reliable choice for automatic pneumonia classification. The findings suggest the potential use of deep learning models, particularly CNN & LSTM, to support medical professionals and the public in quickly and accurately detecting pneumonia through chest X-ray images analysi
Implementation of Transfer Learning on CNN using DenseNet121 and ResNet50 for Brain Tumor Classification
Full text linkBrain tumors are conditions characterized by abnormal cell growth in the brain, which can disrupt brain function. Early detection and accurate classification are crucial to ensuring effective treatment. This study aims to improve the accuracy of brain tumor classification by implementing Convolutional Neural Networks (CNN) using Transfer Learning approaches on DenseNet121 and ResNet50 models. Transfer Learning leverages knowledge from pre-trained models on larger datasets, thereby accelerating the training process and enhancing performance on the brain tumor dataset. The dataset used consists of medical images, including images of brain tumors and images without tumors. The data was divided into two parts, with 80% for training and 20% for validation. This split ensures that the model learns optimally from the training data and is tested on unseen data to objectively evaluate its performance. Experimental results show that the ResNet50 model achieved an accuracy of 98.44% on the validation data, while the DenseNet121 model achieved an accuracy of 96.31%. In conclusion, the ResNet50 model outperformed DenseNet121 in brain tumor classification. The implications of this study demonstrate that the Transfer Learning approach with ResNet50 can serve as an effective tool for automated brain tumor diagnosis, potentially improving patient outcomes through more accurate detection and classificatio
Tsukamoto Fuzzy In IoT-Based Automatic Control System Of Kitchen Smoke MSME Palembang Crackers
Full text linkMSMEs, particularly producers of *kerupuk kemplang* from Palembang, often face challenges in managing kitchen smoke generated during production. This smoke not only pollutes the air and poses health risks to workers but also reduces comfort and productivity. Therefore, this study aims to design an Internet of Things (IoT)-based control system using the Fuzzy Tsukamoto algorithm to automatically regulate the exhaust fan speed based on temperature, smoke concentration, and carbon monoxide (CO) levels. This system introduces technological innovation to enhance efficiency and productivity in MSME kitchen management. The method involves using MQ135, MQ7, and DHT11 sensors to detect kitchen environmental conditions in real time. The collected data is processed by the NodeMCU ESP8266 microcontroller using the Fuzzy Tsukamoto algorithm and is then used to adjust the exhaust fan speed via an AC dimmer. The monitoring results are displayed on the Blynk IoT application for easy access. The study results show that the system successfully reduces smoke concentration by up to 30 ppm and CO levels by 40 ppm while maintaining the kitchen temperature within an optimal range of 49°C to 55°C. With a Mean Absolute Percentage Error (MAPE) of 7.66% and an accuracy rate of 92.34%, the system proves to be effective and responsive to changes in kitchen environmental conditions. The implementation of this Fuzzy Tsukamoto and IoT-based system has a positive impact on improving air quality, ensuring worker health, and increasing MSME productivity. Additionally, this system supports a more modern, efficient, and environmentally friendly kitchen management approach, making it an innovative solution for the *kerupuk kemplang* production industr
Comparison of K-Means and K-Medoids Methods in Clustering High Population Density Areas in Bireuen Regency
No full textThis study examines the population density distribution in Bireuen Regency by applying two clustering algorithms, namely K-Means and K-Medoids, to demographic data from 2019 to 2023. Three main variables were used: total population, number of ID card holders (KTP), and number of households (KK). The clustering results identified three primary groups: very dense, dense, and not dense. Districts such as Kota Juang, Jeumpa, and Peusangan consistently fell into the very dense category, while districts like Pandrah, Gandapura, and Makmur tended to be classified as not dense. Cluster quality was evaluated using the Davies-Bouldin Index (DBI). The evaluation results showed that the K-Means algorithm performed better in most years analyzed, particularly in 2020 with the lowest DBI value of 0.3906. Meanwhile, in 2023, K-Medoids outperformed K-Means, with a DBI value of 0.7724. These findings indicate that K-Means is more effective in handling homogeneous data, whereas K-Medoids is more adaptive to data containing outliers or irregular patterns. Overall, the choice of clustering method depends on the characteristics of the data used. The results provide a spatial overview of population distribution that can support regional planning and data-driven public policy. These findings are expected to serve as a basis for more targeted and equitable regional development planning. For future research, it is recommended to expand the analysis by including additional variables such as area size and socioeconomic indicators, as well as optimizing the number of clusters using methods like the Elbow method or Silhouette Score