Progressive Information, Security, Computer, and Embedded System

Optimization of a FastText-Based BiLSTM Model with IndoBERT Semantic Data Augmentation for Indonesian Text Classification

2026-06-02T11:11:26+00:00

Cognitive assessment through short-answer essays requires a consistent and objective scoring process; however, manual evaluation often suffers from time constraints and inter-rater variability. Automatic Essay Scoring (AES) has emerged as a promising approach to automate the assessment process. This study proposes an optimized Bidirectional Long Short-Term Memory (BiLSTM) model combined with FastText embeddings for Indonesian text classification using semantically augmented data generated by IndoBERT. The training dataset was obtained through the EDA_Synonym_IndoBERT augmentation technique on the UKARA dataset, while the validation and testing datasets consisted of original, non-augmented responses. Model optimization was achieved through the integration of Global Max Pooling to enhance feature representation and class weighting to mitigate class imbalance. Experimental results show that the proposed model achieved an accuracy of 93.49% on the validation set and 78.00% on the independent test set. The performance gap between validation and testing results indicates that, although semantic augmentation increases the diversity of training data, model generalization to previously unseen data remains a challenging issue. Furthermore, the implementation of class weighting improved the model's ability to recognize minority-class instances, achieving a recall score of 92%. These findings demonstrate that architectural optimization and training strategies play a crucial role in improving the performance of Automatic Essay Scoring systems for the Indonesian language

Development of a Project-Based Learning-Based Digital Entrepreneurship Module for JTIK Students

2026-06-02T11:48:29+00:00

This study aimed to develop a Project-Based Learning (PjBL)-based digital entrepreneurship module for students of the Informatics and Computer Engineering Department and to determine the validity and practicality of the developed module. This research employed the Research and Development (R&D) method using the Four-D (4D) development model consisting of the define, design, develop, and disseminate stages. Data were collected through observation, interviews, and questionnaires involving material experts, media experts, and students of the Informatics and Computer Engineering Department, Universitas Negeri Makassar. The results showed that the developed module obtained a material validity score of 98.89% and a media validity score of 98.79%, both categorized as very valid. Furthermore, the practicality test results showed scores of 84.12% in the small-group trial and 86.84% in the large-group trial, which were categorized as very practical. The module was developed using Canva and presented in an interactive flipbook integrated with Quizizz to support more engaging and interactive project-based entrepreneurship learning. Therefore, the developed digital entrepreneurship module is feasible to be used as a learning material in technology-based entrepreneurship learning.

House Door Security Design System Based on Face Recognition on ESP32-CAM

2026-06-03T13:47:02+00:00

Currently, the incidence of theft crimes by breaking into house doors is increasing. The importance of a security system is to prevent unknown parties from stealing or violating privacy without the owner's consent. Biometric technology can create a strong security system, by utilizing the biological characteristics that every human has, such as fingerprints, facial detection, eye retina and voice. One of the biometrics that is considered strong when building a security system is facial recognition. This research uses the Haar Cascade Classifier algorithm supported by OpenCV to increase the accuracy of facial identification based on facial structure and eye feature extraction. The training and testing process is carried out directly (real time) using the OV2640 camera and dataset. The designed prototype consists of an ESP32 CAM microcontroller, relay, and door lock solenoid which is integrated with telegram as notification. Based on the test results, it shows that the accuracy of matching facial images using the Haar Cascade Classifier algorithm which matches the database is 80%. Apart from that, the results of testing the distance of the face to the camera, variations in light, position and facial expressions that can be recognized with the ESP32 CAM camera greatly influence the face detection process. In this case, the effective distance is 25-55 cm in light conditions with a light intensity of 83-450 lux, and the face is facing forward. Apart from that, the system is also able to differentiate between human face objects and non-human face objects. The tool's performance from detection to sending unrecognized image data to Telegram took an average of 6.4 ms. From the test results, it is also known that the perfection of facial appearance that can be recognized with the ESP32 CAM camera has a great influence on the face detection proce

A Comparative Evaluation of Back Translation and Easy Data Augmentation for Indonesian Automatic Short Answer Scoring

2026-06-04T16:37:18+00:00

Automatic Short Answer Scoring (AES) is a Natural Language Processing (NLP) application designed to automatically assess short-answer responses. One of the primary challenges in developing AES systems is the limited size and diversity of available datasets, which can adversely affect a model’s generalization capability. Previous studies have demonstrated that Easy Data Augmentation (EDA) based on IndoBERT-generated synonyms can improve model performance on the UKARA dataset; however, this approach remains limited because the augmentation process is performed at the word level. This study aims to compare the effectiveness of Back Translation and IndoBERT-based Synonym EDA for Indonesian AES systems using the UKARA dataset. To ensure a fair comparison, the dataset, preprocessing procedures, FastText-based text representation, BiLSTM architecture, and evaluation methods were kept consistent across experiments, allowing performance differences to be attributed solely to the augmentation techniques. The experiments were conducted using both Non-K-Fold Evaluation and 3-Fold Cross-Validation scenarios. The results indicate that Back Translation outperformed IndoBERT-based Synonym EDA in most experimental settings, achieving the highest accuracy of 89.00% on Dataset A. Furthermore, the findings suggest that the quality and semantic diversity of the generated data have a greater impact on model performance than merely increasing the amount of training data. Therefore, Back Translation can serve as an effective alternative for enhancing dataset quality and improving the performance of Indonesian AES systems.

Keywords: Automatic Short Answer Scoring, Back Translation, Easy Data Augmentation, IndoBERT, BiLSTM, FastText.

SmartPresence: Android-Based School Attendance Application

2026-06-05T07:02:27+00:00

Student attendance is an important indicator in supporting the effectiveness of the learning process and school administrative management. However, conventional attendance systems that are still conducted manually often lead to various problems, including recording errors, delays in data recapitulation, and low efficiency in attendance monitoring. This study aims to design and develop SmartPresence, an Android-based school attendance application that supports digital, effective, and real-time attendance management. The system was developed using the Waterfall model, which consists of requirement analysis, system design, implementation, testing, and maintenance stages. The application was built using Java programming language in Android Studio, with Firebase Realtime Database as the data storage platform and Quick Response Code (QR Code) technology as the attendance validation mechanism. The results show that SmartPresence successfully integrates user authentication, monitoring dashboard, QR Code-based attendance, attendance history, and digital attendance reports into a single integrated platform. Based on Black Box Testing involving 19 testing scenarios, all system functions operated successfully according to user requirements, achieving a 100% success rate. The findings indicate that SmartPresence is capable of improving the efficiency of attendance management while supporting the digital transformation of school administration through fast, accurate, and easily accessible attendance information.

Keywords: Digital Attendance, Android Application, Student Attendance, School Management, SmartPresence

Detection of AI-Generated Answers in Programming Assignments Based on Automated Grading Systems

2026-07-03T01:59:24+00:00

The use of artificial intelligence (AI) by students in completing programming assignments continues to increase and has created new challenges in maintaining academic integrity. This study examines the patterns and prevalence of AI-generated answers in programming assignments submitted through a Learning Management System (LMS) equipped with automatic assessment features. The LMS used in this study was developed using the ADDIE model and integrated with an AI detection mechanism as the data collection platform. This study involved 109 students from two courses, namely Basic Web Programming and Mobile Programming, at UIN Alauddin Makassar during the Even Semester of the 2026/2027 academic year. From a total of 186 assignment submissions, the system identified 49 submissions, or 26.3%, as answers suspected to have been generated using AI. The analysis covered the distribution of flagging across courses, score comparisons before and after penalties, and the general characteristics of answers indicated as AI-generated. The results show that the level of AI use in Indonesian-language programming assignments is relatively high, with TGS-642 in the Web Programming course recording the highest flagging percentage at 41.5%. This study contributes to understanding the phenomenon of generative AI use in Indonesian programming education and offers practical recommendations for managing academic integrity.

Comparison of Naïve Bayes and C4.5 Algorithms in Classifying the Eligibility of Smart Indonesia Card (KIP) College Scholarship Recipients

2026-08-09T21:31:23+00:00

The Smart Indonesia Card for College Students (KIP Kuliah) program aims to expand access to higher education for students from economically disadvantaged families. However, the scholarship recipient selection process in several institutions still faces challenges because it is conducted manually, requiring considerable time and potentially resulting in inaccurate targeting of eligible recipients. This condition may cause students who meet the eligibility criteria to remain unidentified. This study aims to compare the performance of the Naïve Bayes and Decision Tree C4.5 algorithms in classifying the eligibility of KIP Kuliah scholarship recipients based on student criteria data. The study used a dataset consisting of 76 samples with 8 independent attributes. The analysis was conducted using the Orange application, with 80% of the data used for training and 20% for testing. The performance of both algorithms was evaluated using Area Under the Curve (AUC), Classification Accuracy (CA), F1-Score, Precision, Recall, and Matthews Correlation Coefficient (MCC). The results showed that Naïve Bayes achieved better classification performance than C4.5, with an AUC of 0.857, CA of 0.688, F1-Score of 0.686, and MCC of 0.378. In comparison, C4.5 achieved an AUC of 0.469, CA of 0.500, and MCC of 0.000. These findings indicate that Naïve Bayes is more suitable for classifying the eligibility of KIP Kuliah scholarship recipients within the dataset used in this study and has the potential to support a more efficient and accurately targeted selection process.

Analysis of Minimum Feature Requirements for Lettuce and Weed Classification Based on MobileNetV2 and Support Vector Machine

2026-07-03T01:52:59+00:00

Automating weed removal in lettuce cultivation requires a lightweight and fast computer vision system for implementation on mobile or edge devices. This study aims to analyze the minimum number of features extracted from the MobileNetV2 architecture to accurately classify lettuce and weeds. Features extracted from the global average pooling layer of MobileNetV2 yielded 1,280 base features. The SelectKBest method with Mutual Information criteria was used to reduce the feature dimensionality, followed by classification using a Support Vector Machine (SVM) based on the Radial Basis Function (RBF) kernel. Experimental results showed that the model achieved 100% accuracy using only two minimum features, representing a feature reduction of 99.84%. This feature reduction significantly speeds up computation time, making it ideal for mobile-based computing in the smart agriculture sector.