Jurnal Teknoinfo

Implementation of Integrated Registration System at BBI Using Zoho CRM and ISO 9001:2015

Fri, 09 Jan 2026 16:21:06 +0700

This study aims to improve the registration system at Brilliant Brain Indonesia (BBI) by implementing Zoho CRM, aligned with ISO 9001:2015 Clause 8 requirements. The existing system faced challenges including time-consuming paper-based forms, frequent data entry errors due to manual input, and fragmented data management across two unintegrated systems. Using the Business Process Improvement (BPI) methodology, which includes Organizing for Improvement, Understanding the Process, and Streamlining phases, this research analyzed and optimized the registration workflow. The Wilcoxon signed-rank test was applied to evaluate the statistical significance of improvements before and after system implementation. The customized Zoho CRM platform utilized features such as webforms, workflow automation, automated validation, and real-time analytics. Results showed a significant reduction in registration processing time from an average of 20 minutes 44 seconds to 17 minutes 55 seconds per registration (14% decrease), and a dramatic reduction in data input time from 4 minutes 3 seconds to 4 seconds per record (98% decrease). These improvements enhanced operational efficiency, data accuracy, and compliance with ISO 9001:2015 Clause 8. Although further module development is needed to fully complete documentation processes, this study demonstrates that integrating cloud-based CRM systems with quality management frameworks can substantially improve educational service operations.

Web-Based Library Information System Using ReactJS: Case Study at the Faculty of Mathematics and Natural Sciences, Sam Ratulangi University

Fri, 09 Jan 2026 16:22:24 +0700

Digital transformation has reshaped the library landscape, transitioning from manual to digital systems. This research aims to develop a web-based library information system for the Faculty of Mathematics and Natural Sciences (FMIPA) at Sam Ratulangi University (UNSRAT) using ReactJS, NestJS, and PostgreSQL. The system is designed to address the inefficiencies of manual library management and enhance book information accessibility. The Rapid Application Development (RAD) method was applied, encompassing requirements planning, user design, construction, and cutover. Data were collected through interviews, observations, and surveys. The results demonstrate that the system facilitates online book searching and borrowing, accelerates administrative processes, and improves service efficiency. User satisfaction evaluation through surveys showed high levels of approval. This system is expected to support teaching and learning processes and enhance the quality of library services at FMIPA UNSRAT.

Classification of Green Apple Varieties using Convolutional Neural Network based on RGB Color with MobileNetV2

Fri, 09 Jan 2026 16:23:54 +0700

Manual classification of green apple varieties is often time-consuming, labor-intensive, and prone to human subjectivity. This research aims to develop an automated classification model for green apple types based on RGB color features using Convolutional Neural Network (CNN) with MobileNetV2 architecture. The dataset comprises 1,170 images of three green apple varieties: Golden Delicious, Granny Smith, and Manalagi. Image preprocessing steps include cropping, resizing, background removal, and RGB conversion to enhance feature extraction. The model training and evaluation utilize 5-fold Cross Validation to ensure robustness and generalization. Experimental results demonstrate that the proposed model achieves an average accuracy of 96%, precision of 96.33%, recall of 96.33%, and F1-Score of 96.33%. Furthermore, the model is implemented in a web-based application using the Flask framework to predict apple varieties from input images. Testing on new images shows classification confidence levels of 80.92% for Granny Smith, 87.38% for Manalagi, and 78.43% for Golden Delicious apples. This study confirms that CNN with MobileNetV2 and RGB color features effectively classifies green apple varieties, offering practical implications for agricultural automation and quality control.

Classification of Asphalt Road Damage Based on Images Using the Convolutional Neural Network (CNN) Method

M. Rivan Padila, Arie Qurania, Mulyati Mulyati — Fri, 09 Jan 2026 16:33:07 +0700

Damage to roads can cause inconvenience in driving and can even lead to accidents. Some of the damages that are often found on the road network are such as fine cracks, alligator skin cracks, potholes, asphalt grain release and others. The damage needs preventive handling because it is the main infrastructure in land transportation that is used every day plus areas with very high rainfall such as Indonesia, Damage to the road surface can occur more quickly. One method in artificial intelligence that can be used in identifying damaged roads is Convolutional Neural Networks (CNN). This method is capable of self-learning for object recognition, object extraction and classification and can be applied to high image resolution. The Citra data is taken from the results of google street view mapping with the application of the CNN model using YOLOv5, which is expected to be able to classify images specifically more effectively, objectively and safely in road maintenance efforts later. This research aims to classify image-based asphalt road damage using the Convolution Neural Network (CNN) method. The stages of this research consist of Data Selection, Preprocessing, Data Transformation, Data Mining and Pattern Evaluation using confusion matrix. The results obtained F1 score model of 73.5%, the value of mean Average Precision (mAP) of 75%, this shows that this model is able to classify fairly against all categories of data used.

Integrating Information Systems and Mathematical Models for UI/UX Design in Web-Based Digital Archives

Adina Apriyani, Abdullah Ardi, Mega Wahyu Rhamadani — Fri, 09 Jan 2026 16:53:52 +0700

The rapid growth of digital transformation in public institutions has underscored the urgency of developing effective and user-friendly digital archiving systems. In Indonesia, many government agencies, including the Regional Financial and Asset Management Agency (BKAD) of Kapuas Regency, still face inefficiencies in manual document management, with retrieval times averaging 15–20 minutes per file and high risks of data loss. This study aims to design and evaluate a web-based digital archive system that integrates information systems engineering with mathematical usability assessment, thereby addressing both functional and experiential challenges. The research employed the Design Thinking framework, progressing through empathize, define, ideate, prototype, and testing stages. Prototypes were developed using high-fidelity design tools, while usability evaluations combined subjective and objective measures through the System Usability Scale (SUS), Mission Usability Score (MIUS), and Maze Usability Score (MAUS). The findings demonstrate that the proposed system reduced retrieval times by 90 percent (from 20 minutes to 2 minutes) and achieved an SUS score of 82.5 (Excellent), a MIUS of 76.2 (Good), and a MAUS of 78.6 (Good), all surpassing benchmarks reported in previous studies. These results confirm that combining user-centered design with quantitative evaluation yields reliable outcomes. The study concludes that the hybrid evaluation framework provides both theoretical and practical contributions, while recommending further research on advanced features such as AI-based classification and large.

Analysis of E-Commerce Applications Using the System Usability Scale (SUS) Approach

Tri Wahyudi, Gunawan Budi Sulistyo, Nani Purwati, Noor Hasan — Fri, 09 Jan 2026 00:00:00 +0700

This study aims to measure the level of satisfaction and effectiveness of the application, as well as the obstacles encountered, using the System Usability Scale (SUS) approach. The research employs a quantitative method with a questionnaire based on the System Usability Scale (SUS) to assess application usability. The population consists of Generation Z, with a sample of 411 respondents selected using cluster disproportionate random sampling. Data were analyzed to calculate the SUS score (0–100) and interpreted using the adjective scale. This study compares the usability of three e-commerce applications: Shopee, Tokopedia, and Lazada. The results indicate that the usability scores of the three e-commerce applications are relatively close: Shopee (90.06), Tokopedia (91.98), and Lazada (88.42). Based on gender, females prefer Shopee, while males favor Tokopedia and Lazada. In terms of age, the 20–24 age group is more dominant than the 15–19 age group. Meanwhile, based on occupation, students tend to use Lazada, whereas private employees prefer Shopee and Tokopedia. Shopee and Tokopedia demonstrate optimal performance, while Lazada has development potential, particularly for private employees and users aged 15–19. Further research is recommended to deepen the analysis of Lazada and include other regions and age groups for broader results.

Development of A Telegram Chatbot for Final Project Services Using the Design and Development Research (DDR) Approach

Laisa Nurin Mentari, Tifanny Nabarian, Slamet Santoso, Dymas Sutrysno — Sat, 10 Jan 2026 00:00:00 +0700

Final project administration in higher education involves multiple stakeholders and frequently encounters issues such as delayed information delivery, document inconsistency, and repetitive student inquiries. These problems may increase administrative workload and hinder the final project process. This study aims to analyze user needs and develop a chatbot as an academic information service for final project administration at STT Terpadu Nurul Fikri. The research adopts the Design and Development Research (DDR) approach, focusing on two main stages: Need Analysis and Design and Development. User needs were identified through interviews, observations, and document analysis. Based on the findings, a Telegram-based chatbot was designed and developed using Python and Telegram Bot API integration. The chatbot provides academic etiquette guidelines, general final project information, implementation procedures, schedules, references, and frequently asked questions. The novelty of this study lies in its systematic emphasis on the needs analysis stage as a foundational element for academic chatbot development, which has received limited attention in prior studies. The results indicate that the chatbot has been successfully implemented and deployed as a prototype, serving as a basis for future evaluation and further system enhancement.

Comparison of SMOTE and ADASYN in Optimizing Random Forest Model for Imbalanced Financial Ratio Bankruptcy Prediction

Novanda Rizky Ramadhana, Fuad Muhajirin Farid, Yeni Rahkmawati — Mon, 12 Jan 2026 10:44:58 +0700

Classification is a data analysis process that can predict classes based on predefined characteristics. In the era of big data, classification can be performed using machine learning. The problem of machine learning in classification analysis is imbalance data which often affect model performance. SMOTE and ADASYN are oversampling techniques to solve this problem. This study aims to evaluate the effectiveness of SMOTE and ADASYN in improving the performance of the Random Forest model on imbalanced data in the case of company bankruptcy using financial ratios. Models were built using training data with various splitting data and oversampling techniques. Then, the resulting models will be tested using testing data. The results show that the best model was achieved with a combination of splitting data 70:30 using SMOTE technique, which produced the highest f1-score of 40.57%, compared to ADASYN technique with 36.11% (a decrease of 4.46%), and without oversampling techniques with 19.51% (a decrease of 21.06%). The findings indicate SMOTE and ADASYN can identify minority values which are the main problem of imbalance data, with SMOTE showing better performance compared to ADASYN.

Development of a Plant Weed Detection Model Using the Mask R-CNN Algorithm for Smart Farming

Budi Prayitno, Pritasari Palupiningsih, Atam Rifai Sujiwanto — Mon, 12 Jan 2026 00:00:00 +0700

A more efficient and sustainable agricultural system is urgently needed during world population growth and global climate change. One of the main challenges is that ineffective weed management can significantly reduce crop yields. Conventional farming methods, such as large-scale herbicide application, also negatively impact the environment. Therefore, the development of smart farming technology based on artificial intelligence (AI) is a crucial innovative solution. This research is urgent in the context of developing AI-based systems that significantly contribute to agricultural technology. The urgency of this study is the creation of a plant weed detection model using deep learning to determine the readiness of planting land with high accuracy values. The importance of this research lies not only in the development of technology, but also in its contribution to the farmer economy and the progress of the agricultural sector in Indonesia. This research aims to build and develop a plant weed detection model using deep learning to determine the readiness of planting land, as well as evaluate the detection model built to produce high accuracy. The research method used follows a flow consisting of problem understanding, data understanding, data preprocessing, modelling, and evaluation. The deep learning method used is object detection by applying the Mask R-CNN algorithm with the ResNet-50 architecture as the backbone. The evaluation of model performance was carried out using Mean Average Precision (MAP). The results of this study demonstrated the development of a deep learning-based weed detection model using the Mask R-CNN algorithm, which achieved a MAP of 37.32 and was able to overcome the challenges of varying weed types, lighting conditions, and complex field conditions.