Claim Missing Document
Check
Articles

Found 15 Documents
Search

Optimasi Klasifikasi Daun Jeruk Menggunakan JST dengan GLCM dan LBP Komas Syamsuddin; Herri Setiawan
Jurnal Algoritme Vol 7 No 1 (2026): October 2026 || Algoritme Jurnal Mahasiswa Teknik Informatika
Publisher : Program Studi Teknik Informatika Universitas Multi Data Palembang

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35957/algoritme.v7i1.15770

Abstract

Accurate identification of citrus plant seedlings is essential to improve agricultural quality and productivity; however, manual identification methods remain subjective. This study aims to optimize citrus leaf image classification using a Backpropagation Neural Network (BPNN) through the combination of texture features, namely Gray Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP), as well as the analysis of cropping size variations and edge detection methods. The dataset consists of 10,000 citrus leaf images categorized into five classes. The research stages include image preprocessing, cropping size variations (150×150 to 480×480 pixels), edge detection (Canny, Sobel, and Prewitt), feature extraction, and classification using BPNN. Performance evaluation using accuracy, precision, recall, and F1-score indicates that preprocessing configuration significantly affects model performance. The best result is achieved at a cropping size of 400×400 pixels with Sobel edge detection and combined GLCM+LBP features, yielding an accuracy of 86.2%. In contrast, single feature approaches produce lower accuracy, with 71.4% for GLCM and 74.6% for LBP. These findings demonstrate that the combination of texture features effectively improves the accuracy of citrus leaf identification. Keywords: Neural Network, GLCM, LBP, Edge Detection, Citrus Leaf
Spatiotemporal Prediction of Medical Device Sales Using Hybrid LSTM GNN in South Sumatra Khoirusy Syafaat; Herri Setiawan
Journal of Computer Networks, Architecture and High Performance Computing Vol. 8 No. 4 (2026): Research Paper October 2026
Publisher : Information Technology and Science (ITScience)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47709/cnahpc.v8i4.9538

Abstract

Forecasting medical device purchases at the regional level supports equitable allocation, procurement planning, and distribution scheduling across administrative regions. In developing regions where infrastructure and logistics capacity are unevenly distributed, anticipating regional purchase patterns mitigates supply imbalances and improves the continuity of clinical services. The demand for medical devices exhibits complex spatiotemporal characteristics, where demand variations are influenced by temporal dynamics and interregional spatial connectivity. The objective of this study is to apply and evaluate a Hybrid Long Short-Term Memory and Graph Neural Network (LSTM-GNN) model for predicting the spatiotemporal demand for medical devices, to analyze the effect of spatial relationships on demand variations, and to identify optimal hyperparameter combinations. The methodology uses medical device sales data from PT Parit Panjang, a distribution company in South Sumatra, covering July 2020 to October 2025, comprising 10,878 unique invoices across 16 regencies and cities (Musi Rawas Utara, with only 2 invoices in the study period, was merged into Musi Rawas). Records were aggregated into a daily panel, split chronologically into training, validation, and testing sets (July 2020 to December 2023, January to December 2024, and January to October 2025), and normalized with a log transformation fitted on the training set only. Because 81.54% of all daily region-level targets across the full panel are zero (84.85% in the test period), the problem is formulated as zero-inflated forecasting with two stages: a binary classifier that predicts whether a sale occurs and a regression model that predicts the sales amount on non-zero days. A binary COVID-19 pandemic indicator (July 2020 to December 2022) is included as an exogenous input channel. A spatial graph was built from regional proximity using Haversine distance with three nearest neighbors and a maximum distance threshold of 104.66 km. Temporal dynamics were modeled with LSTM, spatial dependencies with a Graph Convolutional Network (GCN), and both representations were combined through concatenation followed by a fully connected layer. The Hybrid LSTM-GNN achieved an R-squared of 0.0523, an RMSE of 8,002,603 IDR, and a MAE of 1,881,358 IDR on the original scale, outperforming the standalone LSTM (R-squared 0.0433), the standalone GNN (0.0059), and the always-zero baseline (0.0059). The classification stage achieved 90.67% accuracy and an AUC of 0.861. Diebold-Mariano tests confirm that the hybrid model improves over the standalone LSTM and GNN and over the statistical baselines at p < 0.001, but its practical advantage is limited to the classification stage. A random search over 30 configurations identified a superior configuration (LSTM hidden 219, GNN hidden 107, dropout 0.17) reaching an R-squared of 0.0582, the best among all evaluated models. The COVID-19 indicator provides a small improvement, and replacing the distance-based graph with the actual distribution network (a star centered on Palembang) raises the R-squared to 0.0703. In conclusion, the zero-inflated Hybrid LSTM-GNN framework supports regional medical device sales forecasting and distribution planning, with the classification stage providing the dominant predictive signal.
Polynomial Regression Model for Predicting PVC Ceiling Sales Based on Historical Data: A Case Study at CV. Kilaw Maju Bersama M. Ilham Akbar; Herri Setiawan; Zaid Romegar Mair
Jurnal Kecerdasan Buatan dan Teknologi Informasi Vol. 5 No. 3 (2026): September 2026
Publisher : Ninety Media Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.69916/jkbti.v5i3.597

Abstract

The growth of the building materials industry, particularly PVC ceilings, requires accurate sales forecasting to support production stability, distribution efficiency, and inventory management. CV. Kilaw Maju Bersama, a PVC ceiling distributor, faces weekly sales fluctuations based on 104 weekly observations from January 2023 to December 2024. This study develops a Polynomial Regression model to forecast PVC ceiling sales using average selling price, number of retail stores, and promotional activities as independent variables. A quantitative approach was applied through data collection, data cleaning, exploratory data analysis (EDA), chronological time-series train-test splitting without shuffling, Z-score standardization, polynomial degree selection using TimeSeriesSplit, and model evaluation using MAPE, RMSE, and R 2 . The results show that the degree-1 Polynomial Regression model achieved a training MAPE of 3.12%, RMSE of 129.741, and R 2 of 0.643, while the testing set achieved a MAPE of 4.57%, RMSE of 187.311, and R 2 of 0.380. These results indicate low prediction error and satisfactory generalization to unseen data. The model therefore provides a practical decision-support tool for weekly sales forecasting, inventory planning, and data-driven business decision-making in the building materials industry.
KLASIFIKASI JENIS-JENIS SEL DARAH PUTIH METODE MENGGUNAKAN JARINGAN SARAF KONVOLUSIONAL (CNN) Erik Pratama; Herri Setiawan; Mustafa Ramadhan
Jurnal Teknik Informatika dan Komputer Vol. 5 No. 2 (2026): Jurnal Teknik Informatika dan Komputer
Publisher : Universitas Muhammadiyah Prof. DR. HAMKA

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.22236/jutikom.v5i2.23380

Abstract

White blood cells are essential components of the immune system, and manual identification of their subtypes requires expert observation and considerable examination time. This study develops a Convolutional Neural Network (CNN) model based on the VGG19 architecture to classify four white blood cell types: eosinophils, lymphocytes, monocytes, and neutrophils. The dataset consists of 2,400 locally acquired microscopic images processed to a resolution of 224 × 224 pixels. The model was evaluated using two training-validation split ratios, 80:20 and 70:30, and two training durations, 10 and 20 epochs. The best configuration was obtained with the 80:20 split and 20 epochs, achieving 93.94% training accuracy, 88.50% validation accuracy, 0.1692 training loss, and 0.3181 validation loss. The contribution of this study lies in applying VGG19 to locally acquired white blood cell images and systematically evaluating the effects of the training-validation ratio and number of epochs on the same dataset. The results show that the 80:20 split with 20 epochs provides the best validation performance among the evaluated scenarios.
Design of Additional Safety Locks for Motor Vehicles Using RFID Herri Setiawan; Tasmi Tasmi; Dhamayanti Dhamayanti
Journal of Information System and Informatics Vol 5 No 4 (2023): Journal of Information Systems and Informatics
Publisher : Asosiasi Doktor Sistem Informasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51519/journalisi.v5i4.620

Abstract

Indonesia, facing the dual challenges of rapid economic development and rising vehicle theft rates due to economic inequalities, has prompted a need for innovative security solutions in the automotive sector. This research aims to address this growing concern by introducing a sophisticated motor vehicle security system utilizing Radio Frequency Identification (RFID) technology. The proposed system employs RFID keys, which use ID tag cards as unique identifiers to activate motor vehicles. The methodology centers around the integration of RFID technology with a microcontroller, creating a security system that is both advanced and user-friendly. This system includes an alarm feature that is triggered when an ID tag card not registered with the vehicle's RFID is used, providing an additional layer of security against unauthorized access. The results of implementing this system have been promising. Testing revealed that the RFID locks, when combined with the microcontroller, effectively recognized authorized ID tag cards and successfully denied access when unauthorized cards were used. The alarm system also functioned as intended, activating when discrepancies in ID tag data were detected. This integration of RFID with a microcontroller not only enhanced the security of motor vehicles but also contributed to a significant reduction in vehicle thefts in the community, showcasing the potential of RFID technology as a viable solution to the prevalent issue of vehicle security in Indonesia.