METHODIKA: Jurnal Teknik Informatika dan Sistem Informasi

LEXICON BASED ANALISIS DAN RANDOM FOREST TERHADAP ISU POLITIK DINASTI INDONESIA PADA APLIKASI X

Humuntal Rumapea — 2026-03-30

Dynastic politics in Indonesia remains a widely discussed issue, eliciting diverse public opinions ranging from support as a political right to criticism of democratic quality, with social media, particularly the X platform, serving as an important venue for public sentiment analysis. This study employs a combination of the Lexicon Based method using the InSet Lexicon and the Random Forest algorithm to analyze public sentiment on dynastic politics. The dataset consists of 1,593 tweets collected from August 1 to December 24, 2024, which underwent text preprocessing, labeling into three sentiment categories: positive, negative, and neutral, and word weighting using TF-IDF. The methodology includes splitting the data into training and testing sets with an 80:20 ratio, applying undersampling on the training data to balance class distribution, and training a Random Forest model with 100 decision trees and a maximum depth of 5 per tree, based on the entropy criterion. Evaluation results show that the model successfully classifies public sentiment with an accuracy of 89%, precision of 82%, recall of 81%, and f1-score of 81%.

SISTEM KEAMANAN BUKA PINTU OTOMATIS MENGGUNAKAN IDENTIFIKASI WAJAH BERBASIS MIKROKONTROLER ESP32

Ari Prasetio — 2026-03-30

In this modern era, security systems have become one of the most essential aspects of daily life, especially in home
security. The increasing rate of crime, which continues to rise along with technological advancements, makes it
necessary to implement reliable security systems to protect personal assets and privacy. To enhance home security
more effectively, an Automatic Door Security System Using Face Recognition Based on an ESP32-CAM
Microcontroller was developed. This system allows homeowners to feel safer without constant concern about their
home’s security. The system uses the ESP32-CAM microcontroller as its main controller, combined with a solenoid
door lock module and a web browser interface. These components work together to detect and register faces that
are either newly enrolled or already stored in the system. When a registered face is successfully recognized, the
device displays a green notification and automatically unlocks the door through the solenoid lock mechanism.
Additionally, the captured facial data is processed and stored by the ESP32-CAM for future recognition

KLASIFIKASI PENYAKIT DAUN PADI MENGGUNAKAN TRANSFER LEARNING DENGAN ANALISIS PENGARUH VARIASI DIMENSI CITRA PADA KINERJA MODEL

Akhmad Taukhid — 2026-03-30

Penelitian ini berfokus pada deteksi dini penyakit daun padi untuk meningkatkan produktivitas pertanian dan mengurangi kesalahan diagnosis yang sering terjadi pada identifikasi manual. Meskipun berbagai penelitian telah menerapkan deep learning untuk klasifikasi penyakit tanaman, pengaruh resolusi citra terhadap kinerja model klasifikasi penyakit daun padi, khususnya pada skenario data terbatas, masih jarang dikaji secara sistematis. Penelitian ini bertujuan menganalisis kinerja model klasifikasi penyakit daun padi berbasis transfer learning dengan arsitektur VGG16 pada citra beresolusi 224×224 piksel, sekaligus menilai efisiensi proses komputasi pelatihan dan pengujian yang dilakukan. Data yang digunakan berupa 320 citra daun padi dari dataset publik “Daun Padi Sultra (Sulawesi Tenggara)” di Kaggle yang komprehensif menjadi data latih, validasi, dan uji dengan perbandingan 60:20:20. Tahapan penelitian utama meliputi eksplorasi karakteristik dan distribusi data, pra-pemrosesan citra (pengubahan ukuran ke 224×224, normalisasi, dan augmentasi terbatas), serta pembangunan model transfer learning dengan VGG16 sebagai ekstraktor fitur yang membekukan dan kepala klasifikasi kustom. Model dibor menggunakan optimizer Adam dengan mekanisme EarlyStopping dan ModelCheckpoint, kemudian dievaluasi menggunakan akurasi, presisi, recall, F1-score, dan konfusi matriks. Hasil pengujian menunjukkan bahwa model mencapai akurasi uji sebesar 98,44% dengan loss 0,1815, serta nilai rata-rata makro dan rata-rata tertimbang untuk presisi, recall, dan F1-score yang mendekati 0,98 dengan hanya satu kesalahan klasifikasi pada data uji. Proses pelatihan dan penyelesaian dapat diselesaikan dengan beban komputasi yang masih moderat pada lingkungan GPU Google Colab, sehingga konfigurasi VGG16 dengan resolusi 224×224 piksel berpotensi menjadi baseline yang efektif dan efisien untuk klasifikasi penyakit daun padi pada skenario data terbatas.

IMPLEMENTASI METODE DEPTH FIRST SEARCH PADA SISTEM PAKAR KONSULTASI BANTUAN HUKUM CERAI GUGAT

Rabbil Budiman — 2026-03-30

The problem of divorce, especially in Pontianak city, has increased significantly, but community access to legal information and services is often hampered by a lack of legal understanding, creating a gap in justice that can lead to other social problems. Therefore, this research focuses on the development of an expert system that utilizes the depth first search (DFS) method to provide consultation results on legal issues of contested divorce based on the compilation of Islamic law. The research methods include problem identification, data collection through interviews with legal experts followed by design using the waterfall method to ensure the system runs well. The test results show that the depth first search method is able to provide fast and accurate consultation results regarding the problems faced by wives who want to divorce based on the results of testing the accuracy of the system by comparing the results of system diagnosis and experts obtained 80% results so it is concluded that the accuracy of the system made is considered successful.

ALGORITMA RANDOM FOREST UNTUK PREDIKSI STATUS PINJAMAN BERDASARKAN SKOR KREDIT

Hadit Attaufiqqurrohman — 2026-03-30

The rapid development of financial technology has encouraged financial institutions to adopt data-driven credit scoring systems in order to minimize the risk of default. However, many loan eligibility prediction models still face challenges such as data imbalance (class imbalance) and the limited capability of traditional models to capture non-linear relationships among variables. This study aims to develop a loan status prediction model using the Random Forest algorithm combined with the Synthetic Minority Oversampling Technique (SMOTE) and One-Hot Encoding (OHE) to improve model accuracy and generalization capability. The data used in this study are secondary data obtained from the public Kaggle platform, consisting of 45,000 records with 14 demographic and financial attributes. The research method employs a supervised learning approach with several stages, including data acquisition and preprocessing (data cleaning, normalization, encoding, and data balancing), Random Forest model training, and performance evaluation using accuracy, precision, recall, F1-score, and AUC metrics. The results show that the combination of Random Forest, SMOTE, and OHE achieves high predictive performance, with an accuracy of 94.8%, precision of 95.6%, recall of 93.7%, F1-score of 94.6%, and an AUC value of 0.972. The most influential variables in loan status prediction are credit_score, person_income, and loan_amnt. This approach is proven to be effective in addressing data imbalance issues and improving classification accuracy in identifying creditworthy and non-creditworthy borrowers.

ANALISIS SENTIMEN MASYARAKAT PADA MEDIA SOSIAL PLATFORM X TERHADAP PERTAMINA MENGGUNAKAN METODE DECISION TREE

Yksan Naibaho — 2026-03-30

Platform X social media has become a platform for people to express their opinions, including on issues related to Pertamina. This study aims to analyze public sentiment on Platform X towards Pertamina using the ID3 (Iterative Dichotomiser 3) algorithm-based Decision Tree method. The data used are 2,005 tweets collected with the keyword "Pertamina Corruption". The data went through a preprocessing stage which includes case folding, tokenizing, stopword removal, and stemming. Text features were converted into binary representations (Binary Weighting) of 10 main keywords such as 'corruption', 'pertamina', and 'prosecutor'. The ID3 Decision Tree model was built recursively by selecting separator attributes based on the highest Information Gain value. The results showed that the built model had excellent performance. Evaluation on testing data (20% of the total data) produced an accuracy of 98.50%, with a precision value of 97.98%, a recall of 98.50%, and an F1-score of 98.22%. The 5-fold cross-validation results also confirmed the model's stability, with an average accuracy of 98.30% and a low standard deviation (0.0046). The contains_korupsi attribute was identified as the most informative root node in the decision tree structure. The conclusion of this study is that the Decision Tree method with the ID3 algorithm has proven effective and reliable in classifying public sentiment toward Pertamina on Platform X with high accuracy. The results of this analysis are expected to be used by Pertamina in understanding public opinion and formulating more appropriate communication strategies.

ANALISIS KUALITAS PELAYANAN GEREJA UNTUK MENINGKATKAN PARTISIPASI PEMUDA DALAM KEGIATAN GEREJA DENGAN METODE SERVQUAL BERBASIS WEB

Grace Angelina Kacaribu — 2026-03-30

Service quality is a crucial factor in maintaining congregational participation, especially among youth. Changing trends and technological developments have led the younger generation to expect innovative, accessible, and responsive services. Services that fail to meet expectations can decrease their participation in church activities. This study was conducted in the Els Generation youth community of GPdI El-Shaddai Medan with the aim of analyzing the quality of church services using the SERVQUAL method and designing a web-based application to facilitate evaluation. Data were collected through an online questionnaire from 100 respondents, supplemented by interviews and observations. Analysis was conducted on five SERVQUAL dimensions: tangibles, reliability, responsiveness, assurance, and empathy to measure the gap between expectations and perceptions. The results showed that the four main dimensions—Tangibles (-0.10), Reliability (-0.12), Responsiveness (-0.08), and Assurance (-0.07)—had a negative gap, indicating that church services have not fully met youth expectations. Only the Empathy dimension (+0.05) had a positive gap, indicating the church's strength in personal care and attention. The dimension with the largest negative gap was Reliability (-0.12), indicating a need for improvement in aspects of service reliability, such as activity consistency and schedule certainty. The developed web-based application facilitated data collection, gap calculations, and the presentation of analysis results automatically and efficiently.

ANALISIS KINERJA ALGORITMA RSA PADA ENKRIPSI CITRA DIGITAL BERDASARKAN PARAMETER PSNR DAN MSE

Jamaluddin — 2026-03-30

Digital image security is an important issue in this era of increasingly massive multimedia-based data exchange, especially for sensitive information that requires a high level of protection. This study aims to analyze the performance of the Rivest Shamir Adleman (RSA) asymmetric cryptography algorithm in the digital image encryption process based on the Mean Squared Error (MSE) and Peak Signal-to-Noise Ratio (PSNR) parameters, as well as encryption and decryption times. The method used is a quantitative experiment on 30 digital images with varying resolutions (256×256, 512×512, and 1024×1024 pixels) and two RSA key lengths (1024-bit and 2048-bit). The test results show that the MSE value ranges from 0.001068 to 0.002620 and the PSNR value ranges from 75.08 to 78.34 dB, indicating that the decrypted images are of very high quality and close to the original images. However, the computation time increased significantly with increasing resolution and key length, with RSA 2048-bit taking almost twice as long as RSA 1024-bit. These findings show that the RSA algorithm is very effective in maintaining the integrity of digital images, but has limitations in terms of computational time efficiency, especially for high-resolution images. Therefore, a balance between security and performance is needed in practical implementation.