Article Per Year (5 Year)
p-Index From 2021 - 2026
1.456
P-Index
This Author published in this journals
All Journal JURNAL SISTEM INFORMASI BISNIS Sistemasi: Jurnal Sistem Informasi Jurnal AbdiMas Nusa Mandiri Bulletin of Informatics and Data Science JoDENS: Journal of Digital Ecosystem for Natural Sustainability Al'Adzkiya International of Computer Science and Information Technology Journal JUTEKOM - Jurnal Teknologi dan Ilmu Komputer
Masparudin, Masparudin
Universitas Universal
Computer Science & IT Decision Sciences, Operations Research & Management Economics, Econometrics & Finance Education Electrical & Electronics Engineering Energy Engineering Environmental Science Mathematics Physics Social Sciences Other

Published : 10 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 10 Documents
Search

 1 
SISTEM PREDIKSI KUALITAS SANTAN KELAPA MENGGUNAKAN NEAREST MEAN CLASSIFIER (NMC) Masparudin Masparudin; Abdullah Abdullah; Usman Usman
Sistemasi: Jurnal Sistem Informasi Vol 9, No 3 (2020): Sistemasi: Jurnal Sistem Informasi
Publisher : Program Studi Sistem Informasi Fakultas Teknik dan Ilmu Komputer

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (357.5 KB) | DOI: 10.32520/stmsi.v9i3.1015

Abstract

Santan kelapa adalah salah satu bahan pokok yang selalu dijadikan bahan untuk segala jenis makanan. Kualitas menjadi hal yang penting dalam pemilihan santan kelapa. Bagaimanapun juga pengidentifikasian kualitas santan secara manual tidak efisien, hal ini terjadi karena sulitnya membedakan mana santan yang murni dan mana santan yang bercampur dengan air. Tujuan penelitian ini adalah membangun sistem deteksi kualitas santan berdasarkan warna santan. Algoritma klasifikasi yang digunakan adalah nearest mean classifier (NMC),  Metode ini menghitung jarak vektor input citra ke masing-masing mean kelas dari citra latih, jarak terdekat merupakan dasar dalam menentukan hasil dari klasifikasi.  Evaluasi menggunakan metode validasi holdout menggunakan total 135 citra dengan perbandingan 2/3 untuk data sampel dan 1/3 digunakan untuk data uji. Evaluasi dilakukan dengan 3 menggunakan 3 jenis kamera smartphone yaitu kamera 1 Xiaomi Mi 8 Lite, kamera 2 Oppo F7, dan kamera 3 Samsung Galaxy J3 Pro. Pada pengujian kamera pertama memiliki tingkat akurasi tertinggi yaitu 86,66% dibandingkan dengan kamera 2 dengan akurasi 60% dan kamera 3 dengan akurasi 46%.
Development of Apple Fruit Classification System using Convolutional Neural Network (CNN) MobileNet Architecture on Android Platform Masparudin, Masparudin; Fitri, Iskandar; Sumijan, Sumijan
Sistemasi: Jurnal Sistem Informasi Vol 13, No 1 (2024): Sistemasi: Jurnal Sistem Informasi
Publisher : Program Studi Sistem Informasi Fakultas Teknik dan Ilmu Komputer

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32520/stmsi.v13i1.3533

Abstract

In the current digital era, image classification of fruits, particularly apples, has become crucial for various applications, ranging from agriculture to retail. This research focuses on the utilization of Convolutional Neural Network (CNN) with the MobileNet architecture to classify apple fruit images. Using the Python programming language, three models were successfully trained: Model 1 for apple fruit types, Model 2 for apple fruit diseases, and Model 3 for apple fruit ripeness levels. All three models underwent training and validation, with the final results at epoch 10: Model 1 for apple types achieved an accuracy of 100% and a loss of 0.0046, Model 2 for apple diseases achieved an accuracy of 100% and a loss of 0.0075, while Model 3 for apple ripeness levels achieved an accuracy of 99.76% and a loss of 0.0439. Subsequently, these models were tested on an Android device, and there were two testing scenarios. In the first scenario, each model was tested with 15 images individually. The results showed 100% accuracy for Models 1 and 2, while Model 3 achieved a lower accuracy of 86.67%. In the second scenario, all three models were tested simultaneously using 30 test images, resulting in an accuracy of 55.55%. Several factors, such as limitations in the apple image dataset, particularly in the ripeness dataset, object backgrounds, image capture distances, color and texture similarities, as well as lighting quality, influenced the classification outcomes. To enhance future performance, improved data preprocessing and a combination of detection and classification techniques are needed. This research provides valuable insights for researchers and practitioners looking to implement image classification technology in real-world applications.provides valuable insights for researchers and practitioners looking to implement image classification technology in real-world applications.
Prediksi Kelompok Usia Pengguna Netflix Menggunakan Metode Random Forest Berdasarkan Analisis Genre Tontonan dan Perilaku Pengguna Trafin , Abelina Stevie Maria; Masparudin; Febrianti, Eka Lia; syafrinal, ilwan
Journal of Digital Ecosystem for Natural Sustainability Vol 5 No 2 (2025): Desember 2025
Publisher : Fakultas Komputer - Universitas Universal

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.63643/jodens.v5i2.326

Abstract

The accuracy of user demographics, particularly age, on video streaming platforms is often compromised by the widespread practice of shared accounts. This study addresses this challenge by implicitly classifying user age groups (Youth, Young Adult, Adult, Middle-Aged, Senior) based solely on behavioral data, including viewing genre frequency, sentiment analysis of reviews, and expenditure patterns. The core methodology employs a Random Forest Classifier optimized with SMOTE (Synthetic Minority Over-sampling Technique) to mitigate the severe class imbalance present in the dataset. The initial Baseline Model performed poorly, achieving only 40,13% accuracy and failing to identify minority classes. After implementing SMOTE and hyperparameter tuning, the Final Model demonstrated significant improvement, achieving an Accuracy of 79,26%. The engineered feature, Spend per Person, was identified as the most dominant predictor, validating the approach of using economic factors to differentiate genuine individual usage. Crucially, the model showed exceptional reliability in detecting sensitive age segments, such as Youth (F1-Score 0,88) and Seniors (F1-Score 0,75). This research provides an effective data-driven solution for enhancing age-based content personalization and parental control features.
Early Detection of Patient Surge Anomalies in Hospitals: A Comparative Analysis of Gradient Boosting, Random Forest, and SVM Masparudin Masparudin; Marfuah Marfuah; Abdullah Abdullah
Jurnal Sistem Informasi Bisnis Vol 15, No 4 (2025): Volume 15 Number 4 Year 2025 (In Press)
Publisher : Diponegoro University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/vol15iss4pp488-495

Abstract

Unpredictable fluctuations in patient visits often lead to resource unpreparedness and decreased service quality in hospitals. This study aims to develop an early warning system for patient surges across 110 healthcare service units. Unlike conventional approaches utilizing static thresholds, this study proposes a Statistical Anomaly Detection method based on Z-Score for dynamic labeling and applies Synthetic Minority Over-sampling Technique (SMOTE) to address extreme data imbalance. Three classification algorithms—Gradient Boosting Classifier (GBC), Random Forest (RF), and Support Vector Machine (SVM)—were compared using time-series lag features and volatility trends. Experimental results demonstrate that Gradient Boosting outperformed other methods, achieving the highest F1-Score of 37.35% and a Recall of 48.98%. Although the F1-Score reflects the extreme nature of the data imbalance, achieving high recall is explicitly prioritized in healthcare operations to minimize the critical risk of missed surge events. This study concludes that integrating statistical anomaly-based labeling with ensemble boosting algorithms effectively mitigates noise in heterogeneous hospital visit data, thereby serving as a reliable basis for proactive managerial decision-making.
Enhancing Air Traffic Forecasting Accuracy at Hang Nadim Airport Using ARIMA-Neural Network Masparudin Masparudin; Abdullah Abdullah; Raymond Erz Saragih; Yonky Pernando; Ilwan Syafrinal
SISTEMASI Vol 15, No 4 (2026): Sistemasi: Jurnal Sistem Informasi
Publisher : Universitas Islam Indragiri

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32520/stmsi.v15i4.6265

Abstract

Passenger traffic fluctuations at Hang Nadim International Airport exhibit extreme volatility influenced by the unique characteristics of the Free Trade Zone (FTZ). Single statistical methods often fail to capture non-linear patterns in this high-variability data. Therefore, this study proposes a Hybrid ARIMA-Neural Network model to enhance forecasting accuracy. The primary variable used is the total monthly passenger volume (arrivals and departures). The research stages began with data preprocessing (80:20 train-test ratio), linear component modeling using ARIMA, residual extraction, and non-linear component modeling using Multi-Layer Perceptron (MLP) to correct residual errors on a one-step-ahead basis. Evaluation results show that the standalone ARIMA model is slow to anticipate extreme surges, resulting in a Mean Absolute Percentage Error (MAPE) of 23.75%. The hybrid model integration proved successful in compensating for these weaknesses, reducing the MAPE value to 12.51%. This achievement represents a 47.33% error reduction from the baseline. In terms of novelty, this hybrid approach provides a highly reliable computational solution for airport management with dual characteristics (tourism and industry) in mitigating uncertainty in capacity planning.
Lightweight CNNs for Eggplant Leaf Disease Classification on a Balanced Dataset: A Comparative Study of MobileNetV3 and EfficientNetV2-B2 Masparudin Masparudin; Yonky Pernando; Ihsan Verdian; Fazlul Rahman
Bulletin of Informatics and Data Science Vol 5, No 1 (2026): May 2026
Publisher : PDSI

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.61944/bids.v5i1.166

Abstract

Eggplant (Solanum melongena) is a vital agricultural commodity, but its yield is highly vulnerable to foliar diseases. Early and accurate detection using deep learning is essential for effective crop management. However, deploying automated detection in real-world agricultural settings faces two primary challenges: (1) severe classification bias caused by complex background noise and data imbalance, and (2) extreme computational constraints that hinder the deployment of conventional deep learning models on farmers' edge devices. This study presents a robust methodology for classifying four conditions of eggplant leaves (Healthy, Leaf Spot, Mosaic Virus, and Insect Pest) by implementing an automated background removal technique and targeted data augmentation, resulting in a perfectly balanced dataset of 1,400 images. Furthermore, this research conducts a comparative analysis between two distinct categories of lightweight Convolutional Neural Networks (CNNs): MobileNetV3-Large (representing ultra-lightweight architectures with 224x224 input resolution) and EfficientNetV2-B2 (representing medium-lightweight architectures with 260x260 input resolution). The models were evaluated based on their accuracy, loss convergence, and computational efficiency using an 80:20 data split and early stopping callbacks to prevent overfitting. Experimental results demonstrate that both models achieved exceptional performance. EfficientNetV2-B2 exhibited superior stability and precision, achieving a peak validation accuracy of 97.50% and a validation loss of 0.073. Meanwhile, MobileNetV3-Large reached a validation accuracy of 96.07% with significantly faster training iterations. These findings indicate that while EfficientNetV2-B2 is highly recommended for precision-critical agricultural diagnostics, MobileNetV3-Large remains a formidable alternative for deployment on edge devices with extreme computational constraints
PELATIHAN APLIKASI BUDTRACK BERBASIS MOBILE UNTUK MONITORING BATAS WAKTU PENGGUNAAN OBAT DI KOTA TEMBILAHAN KABUPATEN INDRAGIRI HILIR M Marfuah; M Masparudin; A Abdullah; S Samsudin
Jurnal AbdiMas Nusa Mandiri Vol. 8 No. 3 (2026): Periode Juli 2026
Publisher : LPPM Universitas Nusa Mandiri

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33480/abdimas.v8i3.8316

Abstract

Public knowledge about the expiration time of medications is still limited, despite ongoing educational efforts. Many people rely solely on the Expiration Date (ED) on the primary packaging, which indicates the time limit for use after production. However, once the packaging is opened, the time limit no longer refers to the ED, but rather to the Beyond Use Date (BUD), which depends on the type of medication, its form, aroma, and color. The BUDTrack application not only facilitates the monitoring of the medication's expiration time but also educates the public regarding the expiration time based on both the ED and BUD. This aims to minimize losses and enhance awareness and adherence to avoid medication misuse. In addition to increasing public education on medication expiration time, it is also necessary to enhance the digital ecosystem in the health sector in Tembilahan. Therefore, it is important to measure the usability of the BUDTrack application using the System Usability Scale (SUS) method within the community, especially in Tembilahan, to evaluate the sustainability of the application’s implementation. The SUS calculation results from 34 respondents showed an average score of 77.3, indicating that the BUDTrack application falls within the Acceptable range, with a C grade scale and an adjective rating of "Good." Therefore, it is concluded that the BUDTrack application is well-accepted by users. The application has a very good potential for sustainability.
Monitoring Expiration and Beyond-Use Dates of Non-Compounded Drugs Through Rule-Based and Machine Learning Approaches Marfuah Marfuah; Masparudin Masparudin; Raymond Erz Saragih
Al'adzkiya International of Computer Science and Information Technology (AIoCSIT) Journal Vol 7, No 1 (2026)
Publisher : Al'Adzkiya

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55311/aiocsit.v7i1.392

Abstract

Public knowledge regarding the time limit for drug usage is still limited, despite existing educational efforts. Many people rely solely on the Expiration Date (ED) stated on the primary packaging, which indicates the post-production shelf life. However, once the packaging is opened, the usage limit no longer refers to the ED, but instead to the Beyond Use Date (BUD), which depends on the drug type, time, form, aroma, and color. The application of a Rule-Based Expert System, through a series of IF-THEN rules and machine learning. Three methods were used: decision tree (DT), Gaussian Naive Bayes (GBN), and K-Nearest Neighbor (KNN). The results showed that the decision tree method (95%), the GBN method (94%), and the KNN method (81%) can assist in monitoring drug use limits, minimizing losses, and enhancing user awareness and compliance to prevent medication errors.
Evaluating Resampling Methods for Imbalanced Necrosis Classification on CT Scans Akhmad Rezki Purnajaya; Masparudin
Jurnal Teknologi informasi dan Ilmu Komputer Vol. 2 No. 2 (2026): April 2026
Publisher : Nolsatu Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.65258/jutekom.v2.i2.56

Abstract

Necrosis, or body tissue death, occurs when there is insufficient blood flow to the tissue, which can be caused by injury, radiation, or chemicals. One of the main challenges in the automated diagnosis of necrosis is data imbalance in medical datasets, where the number of pathological cases is far less than normal cases. To address this issue, this study implements and evaluates various data sampling techniques, including Random Undersampling (RUS), Random Oversampling (ROS), Combination of Over-Undersampling (COUS), Synthetic Minority Over-sampling Technique (SMOTE), and Tomek Link, then using a Support Vector Machine (SVM) as the classifier. The test results show that the best sampling technique is the Synthetic Minority Over-sampling Technique (SMOTE), which successfully achieved an accuracy of 100% and an Area Under Curve (AUC) of 100%, indicating its significant potential in improving the accuracy of necrosis diagnosis from CT scans.
Model Regresi Linear Berganda Untuk Prediksi Nilai Kualitas Tidur Berdasarkan Gaya Hidup Sidarta David Setia; Masparudin Masparudin; Kaharuddin; Musliadi KH
Journal of Digital Ecosystem for Natural Sustainability Vol 6 No 1 (2026): Juli 2026
Publisher : Fakultas Komputer - Universitas Universal

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.63643/jodens.v6i1.387

Abstract

Sleep disorders significantly impact individuals' quality of life and chronic health. Previous research utilizing the Sleep Health and Lifestyle Dataset predominantly focused on Classification tasks—predicting discrete diagnostic categories such as Insomnia or Sleep Apnea. Although classification accuracy has been high, this method fails to provide a continuous, quantitative assessment of the severity of Quality of Sleep (QOS). This study aims to address this limitation by developing and interpreting a Multiple Linear Regression (MLR) model to predict the numeric Quality of Sleep (QOS) score on a 1-10 scale based on lifestyle, demographic, and biometric factors. The MLR model was applied following pre-processing, which included One-Hot Encoding for categorical variables and the removal of diagnostic variables to prevent data leakage. Evaluation results demonstrate that the model achieved excellent performance, confirmed by a high Coefficient of Determination (R2) of 0.957 and a very low Mean Absolute Error (MAE) of 0.145 units. Quantitative analysis of the regression coefficients identified Sleep Duration as the most dominant positive predictor and Stress Level as the most significant negative predictor of QOS. These findings provide an important contribution in the form of an interpretable mathematical equation, which can be utilized by clinicians to make measurable, evidence-based intervention recommendations, shifting the focus from diagnosis to quantitative management and prevention.
Co-Authors A Abdullah Abdullah Abdullah Abdullah Abdullah Fazlul Rahman Febrianti, Eka Lia Ilwan Syafrinal Iskandar Fitri, Iskandar Kaharuddin M Marfuah Marfuah, Marfuah Musliadi K H Pernando, Yonky Purnajaya, Akhmad Rezki Raymond Erz Saragih S Samsudin Sidarta David Setia Sumijan Sumijan Trafin , Abelina Stevie Maria Usman, Usman Verdian, Ihsan