Jurnal Teknologi Informasi: Jurnal Keilmuan dan Aplikasi Bidang Teknik Informatika

OPTIMASI MODEL DETEKSI ALERGEN PADA PRODUK PANGAN DENGAN ALGORITMA SUPPORT VECTOR MACHINE (SVM) DAN ADAPTIVE BOOSTING (ADABOOST)

2025-07-02T07:03:03+00:00

One important aspect that needs to be considered in food production is food safety. The implementation of this food safety aspect includes food products that avoid contamination of chemical, physical, and biological substances that can be harmful to human health. In the implementation of the Makan Bergizi Gratis (MBG) program, problems were found related to allergies in the recipients of this assistance program. According to the World Health Organization (WHO), food allergies are ranked as the fourth most serious public health problem, and the only effective treatment for allergy sufferers is to avoid foods that contain allergens. Allergens themselves are compounds or food ingredients that cause allergies and/or intolerances. Laboratory tests of food products for allergen testing that are still carried out traditionally require a lot of time and money, making food producers reluctant to carry out product testing. A way to detect allergen content in food products that is easier, more practical, and more accurate is needed. The research conducted aims to build a prediction model that can be used to detect allergen content in food ingredients through the implementation of the Support Vector Machine (SVM) data mining algorithm optimized with the Adaptive Boosting ensemble learning boosting algorithm (AdaBoost). The research conducted obtained a model that produces the most optimal performance, namely SVM optimized with the AdaBoost algorithm with the split validation method.

PEMANFAATAN AUGMENTED REALITY SEBAGAI MEDIA PROMOSI KAMPUS POLITEKNIK MANUFAKTUR NEGERI BANGKA BELITUNG

2025-07-30T05:49:25+00:00

The development of technology today is advancing at a rapid pace, particularly in the field of information technology. One of the latest advanced and modern information technologies, such as Augmented Reality (AR), has been widely used as a medium for creating 3D objects, such as objects and buildings. This study aims to explore the use of AR for creating promotional and campus introduction media that can accurately represent the physical forms of buildings and facilities at the Bangka Belitung State Polytechnic of Manufacturing, a vocational higher education institution in Indonesia specializing in manufacturing. This media is developed as an Android-based application and can be accessed via Virtual Reality. The method used in this study is the Multimedia Development Life Cycle (MDLC) method. The research results show that the feasibility test score obtained was 82.5%, categorized as highly feasible. Based on these results, it can be concluded that the application was successfully implemented using AR technology and functions effectively.

DETEKSI GERAK BERDASARKAN FITUR WAJAH MENGGUNAKAN METODE KANADE LUCAS TOMASI (KLT)

2025-04-09T05:29:42+00:00

Research by utilizing facial recognition features related to image processing and computer vision is used to produce a system that is almost close to the human visual system in general. In image processing, the detection of the movement of the rig is carried out so as to produce detection results. A problem that often occurs in the motion detection process is that every moving object in the video will be detected as a moving object. Therefore, this study will try to detect human face objects from the video data to be detected so that the detection results will later produce the detection of face objects. Every process of observing human facial movements requires a careful pre-process stage, because it is related to the observation of very smooth movements and a very fast duration. At this stage, the detection and tracking of the facial area must always be precise so that the observation of movements made in the facial area can be accurate. The solution offered for facial motion detection is to apply the Canade Lucas Tomasi (KLT) method for tracking each feature point. The performance process of KLT in detecting faces is to track each existing face by looking at the point of facial features, after the system records the features of the face, the system will detect every facial movement in the video. So by using the KLT method, it is hoped that the system can detect facial objects in the video. The results of the study by testing as many as 30 samples of video data in the form of recordings of human motion objects succeeded in detecting facial movements with an accuracy level of 96%, Recal 88% and an accuracy level of 86%.

IMPLEMENTASI DATA MINING DALAM MENENTUKAN TATA LETAK PRODUK MENGGUNAKAN ALGORITMA FP-GROWTH

2025-08-20T06:57:13+00:00

This study analyzes purchasing patterns in minimarkets using the FP-Growth algorithm to optimize product layouts. One year of sales transaction data (103,181 transactions) from UNDIKNAS Mart were analyzed through data cleaning, transformation, and aggregation. The FP-Growth algorithm was applied with minimum support 5%, confidence 80%, and lift >1 thresholds. Results identified strong product associations, particularly between "Aqua 600 ml (Tanggung)" and various snacks, with confidence values of 81-93% and lift >5. Implementing these findings in product arrangement increased sales by 15-20% despite store accessibility limitations. Cross-validation using a decision tree model showed 81.67% accuracy. The findings demonstrate FP-Growth's effectiveness in small-scale transaction data analysis. The research provides practical contributions for retailers to boost sales through data-driven product layout optimization. A limitation is the single-location data scope, suggesting the need for broader subsequent studies. This study offers a data-based approach adoptable by small and medium retail businesses to enhance operational efficiency and profits. The research confirms that data mining techniques can significantly impact retail performance even in constrained environments, providing empirical evidence of FP-Growth's practical utility in real-world minimarket settings. The methodology and findings contribute to the growing literature on data mining applications in small-scale retail operations, offering replicable frameworks for similar business contexts.

SPAM EMAIL CLASSIFICATION USING SUPPORT VECTOR MACHINE (SVM) AND TF-IDF: A CASE STUDY WITH THE TREC 2007 AND ENRON-SPAM DATASETS

2025-08-08T06:22:09+00:00

Spam emails represent a substantial concern within the digital landscape, impeding users with unsolicited communications. This study elucidates the utilization of a Support Vector Machine (SVM) coupled with a TF-IDF Vectorizer for categorizing emails into spam and non-spam classifications. The model was developed utilizing two publicly accessible pre-processed datasets: the TREC 2007 Public Spam Corpus and the Enron-Spam Dataset. By employing the TF-IDF algorithm, which allocates heightened importance to infrequent yet pertinent terms, alongside SVM, renowned for its efficacy in textual classification, the model exhibits remarkable efficacy, achieving an accuracy of 99.04%, a precision of 98.57% and a recall of 99.62%. These findings underscore the model's formidable capacity to discern spam emails while concurrently minimizing false positives accurately. This is critical for real-world applications where authentic emails must not be erroneously categorized as spam. Furthermore, this study elaborates on the justification for the selection of TF-IDF and SVM in the context of spam email classification, in addition to the evaluation outcomes of the model, which align with existing literature, wherein the integration of SVM with TF-IDF has demonstrated substantial performance in spam detection endeavours.

ENSEMBLE MAJORITY VOTING UNTUK ANALISIS SENTIMEN DAN EMOSI PADA KOMENTAR YOUTUBE: STUDI KASUS RESIDENT EVIL 4 REMAKE

2025-07-14T08:37:55+00:00

Currently, social media can be said to be one of the important things in the fields of marketing, broadcasting and entertainment, such as the gaming industry. In this case, Sentiment Analysis and Emotion Detection can be a tool for understanding the public's response and perception of the content presented. One of them is for the game Resident Evil 4 Remake, which was announced on March 24, 2023, and received a lot of public response on various social media platforms such as YouTube, one of which received responses in the form of 7177 comments between June 3 2022 and February 9, 2024. The research methodology used includes data collection methodology and simulation methodology, by combining the Naive Bayes algorithm, SVM and BERT using the Majority Voting method where these algorithms were previously trained using two different datasets which showed Naive Bayes performance with an accuracy of 84%, SVM with 89%, BERT with 93% and the Majority Voting Method with 90% accuracy with training using the Resident Evil 4 Remake dataset. And in training with the Steam Game Review dataset, Naive Bayes and SVM were obtained with an accuracy of 53%, BERT with 66%, and the Majority Voting Method with an accuracy of 57%. The Majority Voting classification model trained on the Resident Evil 4 dataset was used to perform Sentiment Analysis classification on comments from the YouTube video entitled "Resident Evil 4 Remake: Reveal Trailer" from the IGN Channel. The ratio of positive and negative sentiments was 60.2% and 39%. .8% with the frequency of emotions of anger, excitement and anticipation appearing most frequently.

SOLAR TRACKING SYSTEM DENGAN ARDUINO NANO DAN SENSOR LDR: PANEL SURYA BERGERAK MENGIKUTI ARAH CAHAYA

2025-06-22T17:32:12+00:00

This study discusses the design of a single-axis light tracking system based on an Arduino Nano microcontroller. The system is designed to automatically direct solar panels to follow the direction of incoming light using two LDR sensors and one servo motor. The LDRs detect light intensity on the left and right sides, while the Arduino controls the comparison logic to move the servo according to the direction of the dominant light. Functional testing was conducted indoors using artificial light. Observation results indicate that the system responds well to light direction. The servo's movement aligns with changes in light intensity, though there is a response delay of approximately 1–2 seconds. The system also demonstrates stability when light intensity is balanced, indicating that the control logic operates effectively. The system's limitations lie in its horizontal movement range and the fact that it has not been tested outdoors. Nevertheless, this prototype can serve as a foundation for developing a two-axis tracking system or integrating it with the Internet of Things (IoT) for further applications.

MINIMALISASI ATTACK SURFACE PADA SERVER WEB MELALUI PENDEKATAN KEAMANAN BERLAPIS

2025-08-07T03:46:18+00:00

Servers used to run WordPress-based web applications face serious challenges due to a wide attack surface, especially when backend and SSH access are exposed to the public. This study designs and implements an access protection architecture for servers and websites by combining Cloudflare proxy, Virtual Private Network (VPN), and Web Application Firewall (WAF) to mitigate these risks. An experimental method is employed by building a simulated LAMP and WordPress-based infrastructure on two Virtual Private Servers (VPS), where all public traffic is routed through Cloudflare and administrative access is strictly limited to VPN tunnels. The test results show that the proposed architecture effectively eliminates unauthorized access to backend pages and SSH services without disrupting public access to the website. This approach demonstrates that a simple layered defense strategy can be practically applied to enhance server security while providing a protection model that can be replicated for similar infrastructures.

PENERAPAN SPATIAL DURBIN MODEL PADA DATA PENYAKIT MALARIA DI INDONESIA

2025-05-06T06:54:52+00:00

The Spatial Durbin Model (SDM) is a special case of the Spatial Autoregressive (SAR) model, involving the addition of spatial lag effects of both the dependent and independent variables. The parameter estimation used in this study is the maximum likelihood estimator. Parameter estimation for the SDM is performed at each observation location using spatial weighting. The spatial weights are calculated based on queen contiguity and customized contiguity weighting methods. This study aims to obtain the SDM and identify the factors influencing the number of malaria cases in Indonesia in 2023. The Lagrange Multiplier (LM) test indicates that there is a spatial lag in the dependent variable, with the parameter ρ being significant at a significance level of α = 0.1. Based on the results of the SDM analysis, it was found that the factors directly influencing the number of malaria cases in Indonesia in 2023 are the percentage of poor population, number of medical personnel and the percentage of households with access to adequate drinking water services. Meanwhile, the factors that have an indirect or spatial lag effect are the open unemployment rate and the percentage of poor population.

DETEKSI WARNA KULIT DAN REKOMENDASI PALET WARNA BERDASARKAN SEASONAL COLOR MENGGUNAKAN CNN

2025-05-19T07:04:18+00:00

The Personal Color trend has grown significantly, assisting individuals in selecting clothing colors that complement their skin tone. This research aims to develop a system using Convolutional Neural Networks (CNN) to detect skin tones from photos and recommend color palettes based on Seasonal Color Theory. The system categorizes skin tones into the four seasonal types: Winter, Summer, Autumn, and Spring, and provides tailored clothing color suggestions. By applying machine learning, this system offers a personalized solution for clothing selection, enhancing the shopping experience. It aligns with the growing popularity of Personal Color trends, helping users make more confident and informed color choices that suit their individual characteristics.