JICDA: Journal of Informatics, Computer Science, Data Science And Artificial Intelligence

SMART DIET ASSISTANT APPLICATION FEATURING A FOOD DIARY AND MENU RECOMMENDATIONS BASED ON THE DECISION TREE ALGORITHM

2026-07-16T02:40:05+07:00

Body health is strongly influenced by a balanced and regular diet, yet many people still struggle to manage their daily food intake according to their calorie needs and health conditions. This study aims to develop Smart Diet Assistant, an Android-based application that helps users record daily food consumption through a food diary feature and provides healthy menu recommendations using the Decision Tree algorithm. The application was developed using Kotlin with a Jetpack Compose interface, Room Database for local storage, and an MVVM architectural approach.

User data including age, gender, body weight, height, activity level, and medical history are used to calculate daily calorie needs via the Mifflin-St Jeor equation and to determine menu recommendations based on BMI status and health conditions such as diabetes, gastritis, and high cholesterol. The Decision Tree applies a rule-based approach to filter appropriate menus for each user profile.

Testing was conducted through Black Box Testing across two iterations and User Acceptance Testing (UAT) with 27 respondents. All functional scenarios passed in the final iteration. UAT results showed a usability score of 4.63/5 (92.59%) and a motivation impact score of 4.54/5 (90.74%), both categorized as Very Good. These results confirm that Smart Diet Assistant effectively functions as a decision support system helping users monitor calorie intake and receive personalized diet recommendations suited to their health conditions

IMPLEMENTATION OF THE FINITE STATE MACHINE ALGORITHM IN A MOBILE-BASED FOUR-WHEELED VEHICLE DRIVING SIMULATION

2026-01-25T01:30:51+07:00

Driving training conventionally conducted in real-world environments poses significant risks including accidents, vehicle damage, and high operational costs. This research aims to develop a mobile-based four-wheeled vehicle driving simulation application by implementing the Finite State Machine (FSM) algorithm to create a safe, accessible, and educational learning medium. The application was developed using Unity Game Engine with C# programming language, following the Multimedia Development Life Cycle (MDLC) methodology consisting of six stages: concept, design, material collection, assembly, testing, and distribution. The FSM algorithm was implemented to model vehicle behavior through states including Idle, Start Engine, Accelerate, Cruise, Brake, Stop, and Reverse, with transitions controlled based on user input. Testing involved black-box testing for functionality and user testing with 10 respondents using questionnaires. Results showed all functions operated successfully with 95% accuracy, average responsiveness of 40-80 ms per transition, and user satisfaction scores above 4.5 on a Likert scale of 1-5. The application has been successfully distributed through Google Play Store, demonstrating that FSM can be effectively applied to mobile-based driving simulation, producing a logical, responsive, and realistic educational medium that contributes to improving road safety in Indonesia.

IMPLEMENTATION OF DCT (DISCRETE COSINE TRANSFORM) AND MEDIAN CUT ALGORITHMS FOR CONVERTING JPG IMAGE FILES INTO GIF FORMAT

2026-07-10T16:41:19+07:00

Digital image compression plays an important role in reducing storage requirements and improving transmission efficiency while maintaining acceptable visual quality. This study implements the Discrete Cosine Transform (DCT) and Median Cut algorithms to convert JPG images into GIF format through a desktop-based application developed using Python. The proposed method applies DCT to transform image data from the spatial domain into the frequency domain, enabling the reduction of high-frequency coefficients that contribute less to visual perception. The reconstructed image is subsequently processed using the Median Cut algorithm for color quantization to generate an optimized color palette suitable for the GIF format. The application was developed with a graphical user interface (GUI) using Tkinter to facilitate image selection, parameter configuration, processing visualization, and performance evaluation. Experimental testing was conducted using multiple color palette configurations (256, 128, 64, 32, and 8 colors). Performance was evaluated based on file size, compression ratio, reduction size, mean squared error (MSE), and peak signal-to-noise ratio (PSNR). The results demonstrate that the proposed approach effectively reduces image file size while preserving acceptable visual quality. A smaller color palette produces higher compression efficiency but increases reconstruction error, resulting in higher MSE and lower PSNR values. Overall, the combination of DCT and Median Cut provides an effective approach for JPG-to-GIF image conversion by achieving a balance between compression efficiency and image quality.

Keywords : Discrete Cosine Transform, DCT, Median Cut, Image Compression, Color Quantization, GIF, JPG, Python, Digital Image Processing

IMPLEMENTATION OF SEQUENTIAL SEARCH ALGORITHM FOR STREET NAME SEARCH IN MEDAN CITY

2026-06-20T21:48:58+07:00

In the era of increasingly rapid development of information technology, the need for fast and accurate access to information is also increasing, especially in searching for location information and street names in urban areas. One of the problems often faced is the difficulty in obtaining street name information along with administrative data, such as sub-districts, villages, and postal codes, in a structured and easily accessible manner. This study develops a website-based street name search system in Medan City by implementing a sequential search algorithm for the data search process and a Geographic Information System (GIS) for presenting street location information visually. The sequential search algorithm is used to match street names based on keywords entered by users by examining data sequentially, so it can be used on unsorted data. GIS is used to display search results in the form of a digital map so that street locations can be understood more clearly. This system is designed to make it easier for users to search for street name information quickly, simply, and accurately without requiring a complicated manual search process.

IMPLEMENTATION OF HIDDEN MARKOV MODEL WITH VITERBI ALGORITHM FOR DETECTING MEDAN SLANG

2026-06-20T21:46:18+07:00

The increasing use of slang language in digital communication has created challenges in automatic text analysis, particularly for regional slang variations such as Medan slang. This study aims to implement the Hidden Markov Model (HMM) with the Viterbi Algorithm to detect Medan slang words in fictional texts. HMM is used to model probabilistic relationships between words through two hidden states, namely standard language and slang language, while the Viterbi Algorithm determines the most probable sequence of states based on observed words. The dataset consists of training and testing data collected from fictional stories containing Medan slang expressions. The proposed method was implemented in a web-based application that provides slang detection, history management, and dataset administration features. Experimental results show that the HMM-Viterbi approach can successfully identify slang words and classify them according to their linguistic categories. The evaluation indicates that the proposed model performs effectively in detecting Medan slang patterns within textual data. This research contributes to the development of Natural Language Processing applications for regional language analysis and demonstrates the potential of probabilistic sequence modeling for local slang detection.

Keywords : Hidden Markov Model, Viterbi Algorithm, Natural Language Processing, Medan Slang,Text Classification.

BANANA SPECIES CLASSIFICATION USING NEURAL NETWORKS

2026-07-22T13:41:44+07:00

Indonesian bananas represent a strategic horticultural commodity with significant economic potential, yet post-harvest supply chains face persistent challenges due to subjective and labor-intensive manual sorting. To overcome these limitations, this study proposes an automated image-based classification framework evaluating eight Indonesian banana cultivars (Musa spp.). The computational pipeline integrates deep feature extraction via a pre-trained Inception-V3 model with a Multi-Layer Perceptron Artificial Neural Network (ANN) comprising three hidden layers. Evaluated across 1,020 test images using a 10-fold stratified sampling protocol, the model achieved a classification accuracy of 86.7%, a balanced $F_1$-score of 86.7%, and an Area Under the ROC Curve (AUC) of 0.985. Pisang Cavendish achieved the highest class-wise accuracy at 96.9%, whereas misclassifications occurred predominantly between morphologically similar cultivars, such as Pisang Kepok and Pisang Barangan. Overall, the low cross-entropy log loss (0.419) and high Matthews Correlation Coefficient (0.848) confirm strong prediction stability. These findings demonstrate that combining deep visual embeddings with neural networks offers a reliable, high-throughput tool for automated produce grading, supporting quality standardization and digitalization in agribusiness supply chains.

Keywords : Banana cultivar classification, Deep learning, Inception-V3

A A GAMIFIED HABIT TRACKER USING WEIGHTED SCORING TO IMPROVE STUDENT PRODUCTIVITY

2026-02-05T21:25:18+07:00

Student productivity is strongly influenced by consistency in building academic habits. However, digital distractions and low motivation often hinder students from maintaining productive routines. Conventional habit tracker applications tend to be passive and lack strong motivational reinforcement, while gamification-based applications commonly employ uniform scoring systems that fail to reflect task priority. This study aims to develop an Android-based habit tracker application that integrates gamification with a Weighted Scoring algorithm to produce a fair and proportional reward system. The research method used is Research and Development (R&D). The application was developed natively using Kotlin, Jetpack Compose, MVVM architecture, and Jetpack DataStore. Core features include habit management (CRUD), Experience Points (XP) calculation, levels, badges, and daily reset mechanisms. Black Box testing results indicate that all system functionalities operate as expected, while User Acceptance Test (UAT) results show very high usability and perceived motivational impact. This study concludes that gamification integrated with a Weighted Scoring algorithm is effective as a digital intervention to enhance student motivation and productivity.

IMPLEMENTATION OF SHA-256 ALGORITHM AND CACHING MECHANISM IN BACKEND MALWARE DETECTION SYSTEM WITH WEB AND MOBILE INTERFACE SUPPORT

2026-07-16T00:55:53+07:00

INTEGRATION OF MACHINE LEARNING AND AUGMENTED REALITY FOR MALAY CULTURAL HERITAGE

2026-07-10T16:39:36+07:00

Cultural heritage digitization has become an essential strategy for preserving local heritage through the adoption of emerging digital technologies. This study aims to investigate the integration of Machine Learning and Augmented Reality (AR) for preserving and promoting Malay cultural heritage. A descriptive research approach was employed through literature review, field observations, image acquisition, and cultural object documentation. The research data consisted of digital images and descriptive information of Malay cultural heritage objects collected from five cultural sites: Maimun Palace, North Sumatra State Museum, Deli Serdang Museum, Tebing Tinggi Museum, and Batubara Museum. The collected cultural objects were curated and classified according to their visual characteristics and historical significance to establish a structured dataset suitable for Machine Learning-based object recognition and AR integration. The results demonstrate that the proposed data collection and classification process successfully produced a well-organized cultural heritage dataset capable of supporting automatic object recognition and interactive visualization in an AR environment. The integration of Machine Learning and Augmented Reality provides an effective foundation for developing intelligent digital cultural applications that enhance public engagement while supporting the preservation, promotion, and accessibility of Malay cultural heritage.

Keywords : Machine Learning; Augmented Reality; Cultural Heritage; Malay Culture; Cultural Heritage Digitizatio

IMPLEMENTATION OF A WEB-BASED DECISION TREE MODEL FOR CLASSIFICATION OF MOBILE GAME ADDICTION POTENTIAL BASED ON DURATION, FREQUENCY, AND SLEEP PATTERN

2026-06-10T11:46:38+07:00

The rapid growth of mobile gaming has increased playing duration and frequency among users, potentially affecting sleep patterns and leading to addictive behavior. Early identification of mobile game addiction is important to help users recognize unhealthy gaming habits. This study aims to develop a web-based classification system for detecting mobile game addiction potential using the Decision Tree (C4.5) algorithm. The classification process is based on three indicators: playing duration, playing frequency, and sleep patterns. A dataset consisting of 19 active mobile game users was collected and processed using entropy and information gain calculations to generate classification rules and construct a decision tree model. The system classifies users into three categories: mild, moderate, and severe addiction potential. The results showed that Playing Duration produced the highest Information Gain value (1.038) and was selected as the root node of the decision tree. System testing demonstrated that all evaluated testing samples were classified correctly, resulting in an accuracy of 100% for the tested samples. The developed web-based system provides clear and interpretable classification results and can support the early identification of mobile game addiction potential based on user gaming behavior.

POSITION-BASED PLAYER PROFILING USING UNSUPERVISED LEARNING ON 2026 WORLD CUP PERFORMANCE DATA

2026-07-22T18:51:37+07:00

This study employs unsupervised learning to profile football players based on performance data from the 2026 FIFA World Cup, addressing a gap in the literature concerning player categorization in international tournament contexts. The dataset comprises 28 performance metrics from 512 players across 32 national teams. Following data preprocessing, K-Means clustering and principal component analysis were applied to identify natural groupings. The optimal cluster count, determined through elbow method, silhouette score (0.42), Davies-Bouldin index, and Calinski-Harabasz index, was established at k=4. This resulted in four distinct player archetypes: Defensive Anchors, characterized by strong defensive contributions and high passing accuracy; Utility Players, demonstrating versatility and high work rate but lower overall ratings; Star Forwards, exhibiting exceptional offensive output and the highest player ratings; and Supporting Attackers, showing moderate attacking contributions with limited playing time. A cross-tabulation with traditional positions revealed substantial but imperfect alignment, with goalkeepers nearly perfectly classified, while midfielders showed the most dispersed distribution. The findings validate the multidimensional nature of football performance and demonstrate the efficacy of unsupervised learning for objective player evaluation, offering practical applications for talent identification, tactical planning, and player development.

Keywords : Unsupervised Learning; K-Means Clustering; Player Profiling; Football Analytics;2026 World Cup; Performance Data.

Implementation of the Naïve Bayes Algorithm in the Assessment of Competency Examinations for Training Program Participants at the Medan Industrial Training Center

2026-05-10T12:21:26+07:00

The development of information technology has driven digital transformation across various sectors, including education and industrial training. One important aspect is the assessment process of competency examinations for training program participants, which has traditionally been conducted manually, resulting in lengthy processing times, potential subjectivity, and low efficiency. This study aims to design and implement a competency examination assessment system based on the Naïve Bayes algorithm at the Medan Industrial Training Center (Balai Diklat Industri/BDI Medan). The research methodology includes problem identification, literature review, data collection in the form of examination questions and participants’ answer results, data preprocessing, system design, implementation using the Python programming language with a MySQL database, and system performance evaluation. The developed system involves three user roles: administrator, assessor, and participant. Participants complete theoretical examinations in the form of multiple-choice questions, while assessors provide evaluations for interviews and practical examinations. The administrator is responsible for managing data and examination questions, as well as processing assessment results using the Naïve Bayes algorithm. The implementation results indicate that the Naïve Bayes–based assessment system is capable of automatically classifying participants’ examination outcomes into Pass or Fail categories with a good level of accuracy. The system has proven to improve efficiency, accelerate the assessment process, and minimize subjectivity compared to manual methods. The conclusion of this study is that the application of the Naïve Bayes algorithm in a competency examination assessment system can serve as an effective and innovative solution for the digitalization of the evaluation process at BDI Medan. Future research is recommended to further develop the system to support essay-type questions using more advanced algorithmic approaches.

Keywords : Examination Assessment, Naïve Bayes, Machine Learning, Competency