Journal of Artificial Intelligence and Technology Information (JAITI) https://ejournal.techcart-press.com/index.php/jaiti <hr> <table width="100%" bgcolor="#f0f0f0"> <tbody> <tr valign="top"> <td width="21%"><img src="/public/site/images/adminojstcp/New_Cover_JAITI.png"></td> <td width="79%"> <table width="100%" bgcolor="#f0f0f0"> <tbody> <tr valign="top"> <td width="30%">Journal title</td> <td width="5%">:</td> <td align="justify" width="65%"><strong>Journal of Artificial Intelligence and Technology Information</strong></td> </tr> <tr valign="top"> <td width="30%">Abbreviation</td> <td width="5%">:</td> <td align="justify" width="65%"><strong>JAITI</strong></td> </tr> <tr valign="top"> <td width="30%">Publisher</td> <td width="5%">:</td> <td width="65%"><strong>Tech Cart Press</strong></td> </tr> <tr valign="top"> <td width="30%">Editor in Chief</td> <td width="5%">:</td> <td align="justify" width="65%"><strong>Sanriomi Sintaro, M.Kom.</strong></td> </tr> <tr valign="top"> <td width="30%">Period Publish</td> <td width="5%">:</td> <td width="65%"><strong>March</strong>, <strong>June</strong>, <strong>September</strong>, and <strong>December</strong></td> </tr> <tr valign="top"> <td width="30%">ISSN</td> <td width="5%">:</td> <td width="65%"><strong><a title="E-ISSN" href="https://issn.perpusnas.go.id/terbit/detail/20230119380863209" target="_blank" rel="noopener">2985-6396 (Online)</a><br></strong></td> </tr> <tr valign="top"> <td width="30%">P-ISSN</td> <td width="5%">:</td> <td width="65%"><strong><a title="P-ISSN" href="https://issn.perpusnas.go.id/terbit/detail/20230119280869503" target="_blank" rel="noopener">2985-5306 (Print)</a><br></strong></td> </tr> <tr valign="top"> <td width="30%">DOI</td> <td width="5%">:</td> <td width="65%">10.58602 (Prefix - by Crossref)</td> </tr> </tbody> </table> </td> </tr> </tbody> </table> <hr> <p align="justify"><strong>Journal of Artificial Intelligence and Technology Information (JAITI)</strong> is a peer-review journal focusing on Artificial Intelligence and Technology Information issues. <strong>Journal of Artificial Intelligence and Technology Information (JAITI) </strong>invites academics and researchers who do original research in artificial intelligence and technology 1nfromation.<strong>Journal of Artificial Intelligence and Technology Information (JAITI)</strong> is published by <strong>Tech Cart Press</strong> in <strong>March</strong>, <strong>June</strong>, <strong>September</strong>, and <strong>December</strong> every year. <strong>Journal of Artificial Intelligence and Technology Information (JAITI)</strong> accept articles in Bahasa Indonesia and English.</p> <p align="justify"><strong>Journal of Artificial Intelligence and Technology Information (JAITI)</strong> has ISSN&nbsp;<a title="ISSN Online" href="https://issn.brin.go.id/terbit/detail/20230119380863209" target="_blank" rel="noopener"><strong>2985-6396 (Online)</strong></a> in accordance with the letter of Statement Number <strong>29856396/II.7.4/SK.ISSN/02/2023</strong>, and ISSN <a title="ISSN Print" href="https://issn.brin.go.id/terbit/detail/20230119280869503" target="_blank" rel="noopener"><strong>2985-5306 (Print</strong>)</a> in accordance with the letter of Statement Number<strong> 29855306/II.7.4/SK.ISSN/02/2023</strong>.</p> <p align="justify">We are proud to announce that our journal has successfully achieved accreditation from the Kementerian Pendidikan, Kebudayaan, Riset, dan Teknologi with Number: <strong>156/C/C3/KPT/2026</strong> with a <strong>SINTA 4</strong>. This achievement is the result of the dedication and hard work of the editorial team, reviewers, and writers who have contributed to maintaining the quality of the published articles. With this accreditation, we are committed to continuing to improve the quality and relevance of published research, as well as expanding the scientific impact on the national and international scope. Thank you to all parties who have supported the development of our journal. We hope that this journal will continue to be a forum for the publication of high-quality scientific works in the future.</p> <p align="justify"><img src="/public/site/images/pritasari/JAITI.png"></p> PT. Tech Cart Press en-US Journal of Artificial Intelligence and Technology Information (JAITI) 2985-5306 Development of a Web-Based Liturgical Presentation System with Quick Switch and QR Code Using R&D Method https://ejournal.techcart-press.com/index.php/jaiti/article/view/243 <p>Church worship services require a system that can assist in organizing and presenting worship materials in a structured manner so that congregants can follow the service effectively. However, the conventional presentation of liturgy, Bible verses, and songs creates challenges in material management, slide transition flexibility, and limited access for congregants. This study aims to design and implement a web-based liturgical presentation system to improve efficiency and flexibility in managing and delivering worship materials in a single church, namely GMIM Kanaan Likupang Dua. The research employs the Research and Development (R&amp;D) method, which includes requirement analysis, system design using Unified Modeling Language (UML), development, testing, and evaluation. The system is developed using PHP with the Yii Framework 2 and MySQL database, implementing the Model-View-Controller (MVC) architecture. The main features include liturgy management, web-based presentation, a Quick Switch feature for rapid content changes, and QR Code integration that enables congregants to access presentation content through personal devices. Black Box Testing results indicate that all system functions operate properly. Furthermore, User Acceptance Testing (UAT) shows a high level of user acceptance, with 98.75% from multimedia staff and 88.86% from congregants. Therefore, the developed system is considered feasible and effective in supporting more structured, responsive, and inclusive worship services.</p> Juan Sebastian Sijron Lahope Benny Pinontoan Wisard Kalengkongan Mahardika Inra Takaendengan Eliasta Ketaren Edwin Tenda Copyright (c) 2026 Juan Sebastian Sijron Lahope, Benny Pinontoan, Wisard Kalengkongan, Mahardika Inra Takaendengan, Eliasta Ketaren, Edwin Tenda https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 64 75 10.58602/jaiti.v4i2.243 Implementasi Speech Recognition dan Levenshtein Distance pada Aplikasi Pembelajaran Huruf Hijaiyah Berbasis Android https://ejournal.techcart-press.com/index.php/jaiti/article/view/242 <p><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">Pembelajaran huruf hijaiyah merupakan fondasi awal dalam kemampuan membaca Al-Qur'an, namun pada praktiknya proses belajar di lembaga pendidikan nonformal masih banyak bergantung pada metode konvensional yang kurang interaktif dan belum menyediakan evaluasi pelafalan secara langsung. Penelitian ini bertujuan mengembangkan aplikasi pembelajaran huruf hijaiyah berbasis Android yang mengintegrasikan teknologi </span></span><em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">pengenalan suara</span></span></em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;"> dan algoritma </span></span><em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">Levenshtein Distance</span></span></em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;"> sebagai </span></span><em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">pemeriksa ejaan</span></span></em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;"> untuk menilai kesamaan hasil pelafalan pengguna terhadap target bacaan. Metode penelitian menggunakan pendekatan pengembangan perangkat lunak yang mencakup penghapusan masalah, studi literatur, analisis kebutuhan, perancangan sistem, implementasi, dan pengujian. Perancangan sistem dilakukan menggunakan flowchart dan use case diagram. Sistem memproses suara pengguna melalui modul pengenalan suara, mengubah teks menjadi, lalu menghitung nilai </span></span><em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">edit jarak</span></span></em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;"> dan kesamaan untuk menghasilkan klasifikasi benar, mendekati, atau salah. Hasil pengujian menunjukkan bahwa aplikasi mampu menjalankan fungsi utama pembelajaran, menampilkan materi, memutar contoh audio, menerima input suara, dan memberikan umpan balik otomatis. Pengujian sampel terhadap 15 menunjukkan 6 data benar, 6 data mendekati, dan 3 data salah, dengan tingkat akurasi sistem sebesar 90%. Temuan ini menunjukkan bahwa kombinasi </span></span><em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">pengenalan suara</span></span></em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;"> dan </span></span><em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;">Levenshtein Distance</span></span></em><span dir="auto" style="vertical-align: inherit;"><span dir="auto" style="vertical-align: inherit;"> layak diterapkan sebagai media pembelajaran interaktif untuk membantu anak-anak berlatih membaca huruf hijaiyah secara lebih mandiri, terukur, dan menarik.</span></span></p> Maknun Fil Safah Chamdan Mashuri Copyright (c) 2026 Maknun Fil Safah, Chamdan Mashuri https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 76 88 10.58602/jaiti.v4i2.242 Aspect-Based Sentiment Analysis of Public Opinion on the Free Nutritious Meal Program using BERTopic on X https://ejournal.techcart-press.com/index.php/jaiti/article/view/245 <p>This study aims to analyze public opinion on the Free Nutritious Meal (MBG) Program on the X platform using an Aspect-Based Sentiment Analysis (ABSA) approach with BERTopic-based aspect extraction. Unlike previous studies that primarily perform sentiment classification at the overall text level, this study identifies specific aspects within public discussions to provide more fine-grained insights. Twitter data were collected and preprocessed, followed by topic modeling using BERTopic to extract topics that were subsequently defined as aspects. Topic quality was evaluated using topic coherence (c_v) and topic diversity metrics. The modeling process initially produced 36 topics with a coherence score of 0.4446 and a diversity score of 0.8541. After relevance-based selection, 18 topics were retained as aspects, with the coherence score increasing from 0.4446 to 0.5370 and the diversity score increasing from 0.8541 to 0.8611. Sentiment labeling was then performed using the Twitter-XLM-RoBERTa model to determine the distribution of positive, negative, and neutral sentiments across each aspect. The results demonstrate that the proposed ABSA approach with BERTopic-based aspect extraction provides a more structured and insightful mapping of public opinion, enabling the identification of aspects with the highest levels of support and indications of opposition toward the MBG Program. These findings are expected to serve as a basis for consideration in data-driven policy evaluation and support more informed decision-making.</p> Carmen Emanuela Dwiva Lisapaly Luther Alexander Latumakulita Rillya Arundaa Copyright (c) 2026 Carmen Emanuela Dwiva Lisapaly, Luther Alexander Latumakulita, Rillya Arundaa https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 89 104 10.58602/jaiti.v4i2.245 Klasifikasi Gerakan Tari Bali Perempuan Menggunakan Metode Spatial-Temporal Graph Convolutional Network (ST-GCN) https://ejournal.techcart-press.com/index.php/jaiti/article/view/253 <p>Pengenalan gerakan tari Bali berpotensi mendukung dokumentasi warisan budaya, media pembelajaran, dan sistem umpan balik gerak berbasis komputer. Namun, model klasifikasi video RGB dapat mempelajari latar belakang, kostum, pencahayaan, atau identitas penari sebagai shortcut, bukan struktur gerakan. Penelitian ini bertujuan menyusun baseline Spatial-Temporal Graph Convolutional Network (ST-GCN) berbasis skeleton untuk klasifikasi 13 gerakan dasar tari Bali perempuan menggunakan MediaPipe Pose. Setiap video diproses menjadi 33 landmark tubuh dengan kanal koordinat x, koordinat y, dan visibility, kemudian distandarkan menjadi 64 frame. Folder train dan validation asli digabungkan hanya untuk validasi silang 5-fold berstratifikasi, sedangkan folder test resmi dipertahankan sebagai holdout akhir. Model menggunakan graf 33 landmark MediaPipe, backbone ST-GCN, dan GCNHead dengan global pooling serta linear classifier. Hasil validasi silang memperoleh top-1 accuracy 99,55% +/- 0,49%, top-5 accuracy 99,92% +/- 0,17%, dan macro F1 99,49% +/- 0,55%. Evaluasi holdout akhir satu kali menghasilkan top-1 accuracy 99,39%, top-5 accuracy 100,00%, dan macro F1 99,39%. Audit duplikasi identifier dan overlap SHA-256 tidak menemukan kebocoran data train-test. Strategi evaluasi ini menegaskan pemisahan antara validasi model dan pengujian akhir. Hasil ini menunjukkan bahwa ST-GCN berbasis skeleton menjadi baseline within-dataset yang kuat untuk pengenalan gerak tari Bali, meskipun generalisasi subject-independent belum dapat diklaim karena dataset tidak menyediakan identitas penari.</p> Daniel Sande Bona Copyright (c) 2026 Daniel Sande Bona https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 105 119 10.58602/jaiti.v4i2.253 Penerapan Metode K-Means Clustering dan Principal Component Analysis (PCA) untuk Pengelompokan Provinsi di Indonesia Berdasarkan Indikator Pendidikan https://ejournal.techcart-press.com/index.php/jaiti/article/view/256 <p>Penelitian ini bertujuan untuk mengategorikan provinsi-provinsi di Indonesia berdasarkan indikator-indikator pendidikan, dengan harapan memberikan gambaran menyeluruh dan berbasis data mengenai kondisi pendidikan. Latar belakang studi ini beranjak dari adanya kesenjangan tingkat pendidikan di berbagai wilayah yang membutuhkan analisis mendalam dengan pendekatan data mining. Data yang digunakan diperoleh dari Badan Pusat Statistik (BPS) yang mencakup sejumlah indikator pendidikan, seperti rata-rata lama bersekolah, angka partisipasi kasar (APK), angka partisipasi murni (APM), angka partisipasi sekolahan (APS), serta persentase penduduk yang tidak pernah atau belum mengenyam pendidikan. Pendekatan yang diterapkan dalam penelitian ini adalah Principal Component Analysis (PCA) untuk reduksi dimensi dan K-Means Clustering untuk pengelompokan data. Langkah-langkah penelitian mencakup preprocessing data, normalisasi dengan menggunakan StandardScaler, reduksi dimensi melalui PCA, dan clustering dengan K-Means. Temuan dari penelitian ini menunjukkan bahwa dua komponen utama dari PCA mampu menerangkan hingga 79% variasi dalam data, sehingga bermanfaat dalam menyederhanakan dataset yang ada. Proses pengelompokan menghasilkan tiga kelompok provinsi dengan karakteristik pendidikan yang bervariasi, yaitu kategori tinggi, menengah, dan rendah. Penilaian menggunakan silhouette score mengindikasikan bahwa model ini memiliki kualitas pengelompokan yang baik. Temuan dari penelitian ini diharapkan dapat memberikan kontribusi dalam membantu pengambilan kebijakan pendidikan yang lebih tepat serta menjadi landasan untuk penelitian selanjutnya dengan menambahkan variabel yang lebih luas.</p> Clinton Lumbantoruan Dyah Ayu Megawaty Copyright (c) 2026 Clinton Lumbantoruan, Dyah Ayu Megawaty https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 120 131 10.58602/jaiti.v4i2.256 Pemilihan Fitur Berbasis Algoritma Metaheuristic untuk Meningkatkan Klasifikasi Tingkat Kesehatan Masyarakat di Pulau Jawa https://ejournal.techcart-press.com/index.php/jaiti/article/view/251 <p>Kompleksitas data kesehatan masyarakat di Pulau Jawa menuntut pemilihan fitur yang tepat agar model prediksi lebih akurat dan efisien. Penelitian ini bertujuan mengembangkan model prediksi tingkat kesehatan masyarakat dengan mengoptimalkan seleksi fitur menggunakan dua algoritma metaheuristik, yaitu&nbsp;<em>Particle Swarm Optimization</em>&nbsp;(PSO) dan&nbsp;<em>Genetic Algorithm</em>&nbsp;(GA). Dataset yang digunakan mencakup 16 indikator sosial, ekonomi, dan kesehatan dari 38 wilayah (29 kabupaten dan 9 kota) di Pulau Jawa tahun 2020. Model&nbsp;<em>Random Forest</em>&nbsp;diterapkan untuk mengevaluasi setiap subset fitur. Hasil menunjukkan bahwa PSO menghasilkan akurasi pengujian tertinggi sebesar 79,1% dengan memilih 7 fitur, mengungguli GA yang mencapai 78% dengan 10 fitur. Fitur-fitur yang secara konsisten terpilih dan berpengaruh signifikan adalah angka harapan hidup, rasio puskesmas, akses air minum layak, pemberian ASI eksklusif, balita gizi kurang, dan angka kesembuhan Covid-19. Temuan ini membuktikan bahwa PSO lebih efektif dalam seleksi fitur dan mampu memberikan dasar pengambilan kebijakan kesehatan berbasis data.</p> Khoirunnisa Afandi M. Habibullah Arief Muhammad Andryan Wahyu Saputra Hikmatul Kamila Copyright (c) 2026 Khoirunnisa Afandi, M. Habibullah Arief, Muhammad Andryan Wahyu Saputra, Hikmatul Kamila https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 132 142 10.58602/jaiti.v4i2.251 Implementasi Algoritma Fisher-Yates Shuffle pada Game Edukasi Pengenalan Buah dan Sayur Berbasis Android https://ejournal.techcart-press.com/index.php/jaiti/article/view/252 <p>Pendidikan Anak Usia Dini (PAUD) merupakan masa keemasan yang krusial untuk stimulasi kognitif, termasuk pengenalan pola hidup sehat. Namun, pembelajaran di TK Darul Falah masih menggunakan metode konvensional yang monoton, sehingga menurunkan minat siswa dalam mengenal jenis buah dan sayuran. Banyak aplikasi edukasi yang tersedia saat ini memiliki kelemahan pada alur yang statis, sehingga anak cenderung menghafal urutan jawaban daripada memahami substansi materi. Penelitian ini bertujuan untuk mengembangkan game edukasi berbasis Android menggunakan metode Design Thinking yang meliputi tahap <em>empathize, define, ideate, prototype</em>, dan <em>testing</em>. Penelitian ini integrasi algoritma Fisher-Yates Shuffle untuk sistem pengacakan soal otomatis yang memastikan setiap sesi permainan memberikan tantangan dinamis dan permutasi urutan yang adil (unbiased). Hasil pengujian fungsional melalui <em>Black-box testing</em> menunjukkan seluruh fitur berjalan valid. Pengujian pengalaman pengguna menggunakan <em>System Usability Scale</em> (SUS) menghasilkan skor rata-rata 91,59, yang menempatkan aplikasi pada kategori <em>Acceptable</em> dengan predikat <em>Best Imaginable</em>. Dapat disimpulkan bahwa penerapan algoritma <em>Fisher-Yates Shuffle</em> efektif mengatasi pola hafalan pada anak dan layak digunakan sebagai media pembelajaran interaktif yang adaptif.</p> Muhammad Akhyarul Fanani Ginanjar Setyo Permadi Copyright (c) 2026 Muhammad Akhyarul Fanani, Ginanjar Setyo Permadi https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 143 154 10.58602/jaiti.v4i2.252 Sistem Pendukung Keputusan Penerimaan Staff Keuangan Menggunakan Pembobotan ITARA dan Perangkingan SMART https://ejournal.techcart-press.com/index.php/jaiti/article/view/249 <p>Penelitian ini bertujuan untuk membangun Sistem Pendukung Keputusan (SPK) dalam proses penerimaan staff keuangan dengan mengintegrasikan metode ITARA sebagai teknik pembobotan kriteria dan metode SMART sebagai metode perangkingan alternatif. Permasalahan utama dalam proses rekrutmen terletak pada subjektivitas penilaian serta kesulitan dalam menentukan tingkat kepentingan setiap kriteria secara objektif. Dalam penelitian ini, kriteria yang digunakan meliputi ketelitian, komunikasi, kemampuan akuntansi, hasil tes, dan wawancara yang dianggap merepresentasikan kompetensi utama kandidat. Metode ITARA diterapkan untuk menghasilkan bobot kriteria berdasarkan variasi data yang signifikan dengan mempertimbangkan ambang ketidakpedulian, sehingga bobot yang dihasilkan lebih mencerminkan kondisi aktual. Selanjutnya, metode SMART digunakan untuk menghitung nilai utilitas setiap kandidat dan menghasilkan nilai preferensi akhir sebagai dasar dalam proses perangkingan. Hasil penelitian menunjukkan bahwa sistem yang dikembangkan mampu memberikan urutan kandidat secara objektif, terstruktur, dan mudah dipahami, sehingga membantu pengambil keputusan dalam menentukan kandidat terbaik secara lebih akurat. Dengan demikian, integrasi metode ITARA dan SMART terbukti efektif dalam meningkatkan kualitas, transparansi, dan konsistensi dalam proses seleksi staff keuangan.</p> Indra Al Rasyid Rohmat Indra Borman Copyright (c) 2026 Indra Al Rasyid, Rohmat Indra Borman https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 155 170 10.58602/jaiti.v4i2.249 Automatic Dewey Decimal Classification of Indonesian Book Metadata Using IndoBERT with Weighted Loss and Context Enhancement https://ejournal.techcart-press.com/index.php/jaiti/article/view/258 <p>This study proposes an automatic Dewey Decimal Classification (DDC) classification framework for Indonesian book metadata by integrating the IndoBERT model strengthened through weighted loss and context enhancement mechanisms. The current escalation of digital book collections poses significant challenges in classification efficiency and information retrieval, while the manual DDC classification process still relies on librarian expertise and is relatively time-consuming. The dataset used includes 2,516 book metadata obtained through the Google Books API and mapped into 14 DDC categories. The context enhancement strategy is implemented by integrating book titles and descriptions into a single text representation, while weighted cross-entropy loss, random oversampling, and simple data augmentation techniques are applied to address class imbalance issues. Model performance is evaluated based on accuracy, precision, recall, and F1-score metrics. Experimental results show that the proposed approach achieves an accuracy of 90.14% and a weighted F1-score of 90.15%, outperforming the baseline IndoBERT model, which only achieved an accuracy of 47.82% and a weighted F1-score of 47.06%. These findings indicate that the combination of weighted loss and contextual text representation can improve the semantic understanding of book metadata while reducing bias towards the majority class in Transformer-based DDC classification.</p> Joko Purwanto Fajar Mahardika Adlan Nugroho Copyright (c) 2026 Joko Purwanto, Fajar Mahardika, Adlan Nugroho https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 171 181 10.58602/jaiti.v4i2.258 Sistem Rekomendasi Parfum Lokal Berbasis Web Menggunakan Content-Based Filtering dan Cosine Similarity https://ejournal.techcart-press.com/index.php/jaiti/article/view/257 <p>Industri parfum lokal saat ini mengalami pertumbuhan yang pesat, namun pada praktiknya hal ini memicu fenomena <em>information overload</em> bagi konsumen yang sering kesulitan membedakan varian aroma karena proses pemilihan di platform digital masih bergantung pada deskripsi tekstual yang kurang interaktif dan belum menyediakan pencocokan preferensi secara personal. Penelitian ini bertujuan mengembangkan sistem rekomendasi parfum lokal berbasis web yang mengintegrasikan metode <em>Content-Based Filtering</em> dengan teknik pembobotan <em>Term Frequency-Inverse Document Frequency</em> (TF-IDF) dan algoritma <em>Cosine Similarity</em> untuk menilai kemiripan kriteria pengguna terhadap target produk dalam basis data. Metode penelitian menggunakan pendekatan pengembangan perangkat lunak yang mencakup identifikasi masalah, studi literatur, pengumpulan data, perancangan sistem, implementasi, dan pengujian. Perancangan sistem dilakukan menggunakan <em>flowchart</em>, <em>use case diagram</em>, dan <em>entity relationship diagram</em>. Sistem memproses masukan kriteria dari pengguna meliputi aroma, <em>occasion</em>, gender, harga, dan konsentrasi, mengekstraksinya menjadi vektor bobot, lalu menghitung nilai jarak (<em>similarity</em>) untuk menghasilkan klasifikasi peringkat produk. Hasil pengujian menunjukkan bahwa aplikasi mampu menjalankan fungsi utama pencarian, menampilkan katalog, menerima input preferensi, dan memberikan umpan balik rekomendasi otomatis secara akurat. Pengujian presisi terhadap 20 sampel skenario pencarian menunjukkan 17 data relevan (<em>True Positive</em>) dan 3 data tidak relevan (<em>False Positive</em>), dengan tingkat akurasi sistem sebesar 85%. Temuan ini menunjukkan bahwa kombinasi TF-IDF dan <em>Cosine Similarity</em> layak diterapkan sebagai mesin rekomendasi interaktif untuk membantu konsumen memilih parfum secara lebih mandiri, terukur, dan akurat, serta meminimalisir risiko kesalahan pembelian (<em>blind buy</em>).</p> Mochamad Wahyu Sultan Fatahila Tanhella Zein Vitadiar Copyright (c) 2026 Mochamad Wahyu Sultan Fatahila, Tanhella Zein Vitadiar https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 182 193 10.58602/jaiti.v4i2.257 Development of a Web-Based First-Person Game of the Legend of Toar and Lumimuut Using Three.js https://ejournal.techcart-press.com/index.php/jaiti/article/view/261 <p>The legend of Toar and Lumimuut is a foundational Minahasan narrative that is increasingly unfamiliar to younger digital audiences. This research develops Toar &amp; Lumimuut: Legend of Minahasa, a browser-based first-person puzzle adventure game that adapts the legend into an interactive 3D experience using Three.js. The study applied an iterative Game Development Life Cycle consisting of initialization, pre-production, production, alpha testing, beta testing, and release. The game contains two narrative levels: the Coast of Mount Wulur Mahatus and Mount Lolombulan. Each level integrates puzzle mechanics with story progression, including sacred torch activation, stone pillar sequencing, prophecy fragment ordering, sacred seed planting, and eternal flame activation. Technical implementation includes procedural terrain generated with Perlin noise, animated GLB characters and props, an NPC dialogue and quest system, inventory management, bilingual English-Indonesian text support, and browser-based deployment without installation. Functional validation used black box testing with 68 test cases covering movement, interface controls, dialogue, puzzles, timers, game-over states, and level transitions. All test cases passed, producing a 100% functional success rate. User acceptance testing with 15 respondents aged 19 to 21 produced an overall score of 84.44%, categorized as very good. Compatibility testing on Google Chrome, Mozilla Firefox, and Microsoft Edge showed that the game remained playable across three laptops without dedicated GPUs. The results indicate that Three.js can support accessible cultural game development while preserving local folklore through meaningful interactive gameplay.</p> Delon Daniel Wolayan Benny Pinontoan Edwin Tenda Stephano Caesar Wenston Ngangi Mahardika Inra Takaendengan Dodisutarma Lapihu Copyright (c) 2026 Delon Daniel Wolayan, Benny Pinontoan, Edwin Tenda, Stephano Caesar Wenston Ngangi, Mahardika Inra Takaendengan, Dodisutarma Lapihu https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 194 202 10.58602/jaiti.v4i2.261 Comparative Analysis of Hybrid ARIMA-LSTM against Statistical and Machine Learning Benchmarks for Commodity Stock https://ejournal.techcart-press.com/index.php/jaiti/article/view/260 <p>Predicting stock prices in Indonesia’s commodities and energy sectors is a complex challenge due to high volatility influenced by global market dynamics and macroeconomic factors. This study aims to test the robustness of the ARIMA-LSTM hybrid model in predicting closing stock prices for six major issuers: ADRO, PTBA, MEDC, ANTM, MDKA, and AALI. The proposed approach employs a dual-input strategy that integrates 27 technical indicators with the linear residuals from the ARIMA model. The research methodology begins with data decomposition using the ARIMA model to capture linear components, followed by modeling the residuals using Long Short-Term Memory (LSTM) to capture complex non-linear patterns. The experimental results show that the hybrid model consistently delivers the best performance compared to single models such as ARIMA, Random Forest, and Single LSTM across all test datasets. In the 1-step-ahead scenario, the hybrid model achieved the lowest average MAPE of 2.20%, while in the 5-step-ahead scenario, the error rate remained at 3.98%. A key finding of this research is the hybrid architecture’s ability to mitigate the extreme overfitting experienced by the Single LSTM model, while providing better prediction stability against variations in issuer characteristics. This study concludes that the integration of statistical decomposition and deep learning provides a reliable framework for investors and analysts to make data-driven decisions amid the volatile fluctuations of the Indonesian capital market.</p> Muhammad Iszul Wilsa Heru Purnomo Kurniawan Rizki Dewantara Dinda Febrihastatiwi Indri Setiawati Copyright (c) 2026 Muhammad Iszul Wilsa, Heru Purnomo Kurniawan, Rizki Dewantara, Dinda Febrihastatiwi, Indri Setiawati https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 203 216 10.58602/jaiti.v4i2.260 Web-based E-Commerce Application for MSME in Manado https://ejournal.techcart-press.com/index.php/jaiti/article/view/263 <p>Digital transformation has opened new opportunities for Micro, Small, and Medium Enterprises (MSMEs) in Manado City to increase market reach and strengthen business competitiveness through online transactions. However, many local MSMEs still have limited access to independent digital platforms, while existing e-commerce systems do not always provide the flexibility needed to handle both physical products and service-based offerings in a localized context. This study aims to design and implement a web-based MSME marketplace for Manado City using the Waterfall software development method. The system was developed with a decoupled architecture consisting of Next.js as the frontend framework, Strapi Headless CMS for backend management, and PostgreSQL as the database. The main features include seller store registration, product and service catalog management, dynamic product variants, map-based location pinning, Live Commerce, Midtrans payment gateway integration, order tracking, and accessibility support. System evaluation was conducted using Black-Box Testing and User Acceptance Testing (UAT) with a five-point Likert scale involving buyer and seller respondents. The Black-Box Testing results show that all tested features functioned as expected. The UAT results produced an average acceptance rate of 89.88% from buyers and 91.38% from sellers. These results indicate that the proposed marketplace is functional, practical, and well accepted by users, making it a suitable digitalization solution for supporting MSMEs in Manado City.</p> Gabriell Cristiano Agung Taroreh Benny Pinontoan Edwin Tenda Christian Soewoeh Wisard Kalengkongan Dodisutarma Lapihu Copyright (c) 2026 Gabriell Cristiano Agung Taroreh, Benny Pinontoan, Edwin Tenda, Christian Soewoeh, Wisard Kalengkongan, Dodisutarma Lapihu https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 217 235 10.58602/jaiti.v4i2.263 Performance Analysis of ECS Architecture in 2D Mobile Game Development: Ocean Hero https://ejournal.techcart-press.com/index.php/jaiti/article/view/267 <p>Mobile game development frequently encounters computational performance bottlenecks when a system must render and update the logic of many objects simultaneously in each frame. Conventional Object-Oriented Programming (OOP) architecture produces high memory overhead and elevated cache miss rates because game objects are allocated in scattered, non-contiguous memory locations. This research aims to design, implement, and analyze the performance of an Entity Component System (ECS) architecture in a 2D Android educational arcade game titled Ocean Hero. The development process followed the Game Development Life Cycle (GDLC). ECS separates identity, data, and behavior into entities, components, and systems, allowing game logic to process homogeneous component data sequentially through Unity DOTS. Evaluation was conducted on a Samsung Galaxy A15 4G by comparing ECS and OOP implementations through white-box functional verification and stress testing across six entity workloads from 500 to 3,000 entities, each observed over a 20-second tracking period. The ECS implementation maintained a stable 30 FPS and 33.3 ms frame time across all tested entity levels. In contrast, the OOP implementation degraded to 11 FPS and 90.73 ms frame time at 3,000 entities. Based on the relative performance improvement formula, ECS achieved approximately 172.7% higher runtime performance than OOP at the highest workload. These results confirm that ECS is an effective architectural solution for improving scalability and computational efficiency in real-time 2D mobile games with large entity counts.</p> Raynaldi Irfansya Regar Benny Pinontoan Christian A. J. Soewoeh Copyright (c) 2026 Raynaldi Irfansya Regar, Benny Pinontoan, Christian A. J. Soewoeh https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 236 244 10.58602/jaiti.v4i2.267 Enhancing Sentiment Classification Performance on Tentang Anak Application Reviews Using Optimized Support Vector Machine https://ejournal.techcart-press.com/index.php/jaiti/article/view/271 <p>The increasing use of parenting and child development applications has generated a large volume of user reviews containing valuable insights regarding application quality, usability, and user satisfaction. One of the widely used applications in Indonesia is Tentang Anak: Kehamilan &amp; Anak. However, manually analyzing these reviews is inefficient due to the large amount of unstructured textual data. Therefore, this study aims to enhance sentiment classification performance on user reviews of the Tentang Anak: Kehamilan &amp; Anak application using an optimized Support Vector Machine (SVM) model. The dataset consisted of user reviews collected from application platforms, which were processed through several text preprocessing stages, including cleaning, normalization, tokenization, stopword removal, and stemming. Sentiment labeling was conducted using polarity scores to classify reviews into positive and negative sentiments. The proposed model was evaluated using different test size scenarios (0.1, 0.2, 0.3, and 0.4) and random state configurations to identify the optimal parameter setting. Experimental results demonstrate that the best performance was achieved at a test size of 0.1 with random state 0, obtaining an accuracy of 89.8%, precision of 91.7%, recall of 55.0%, and F1-score of 68.8%. The findings indicate that the optimized SVM model is effective in classifying sentiment in reviews of the Tentang Anak: Kehamilan &amp; Anak application, particularly in achieving high precision and classification stability across multiple testing scenarios. Furthermore, the study highlights the importance of parameter optimization in improving sentiment analysis performance for user-generated textual data.</p> Riska Aryanti Eka Fitriani Royadi Royadi Dian Ardiansyah Copyright (c) 2026 Riska Aryanti, Eka Fitriani, Royadi Royadi, Dian Ardiansyah https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 245 258 10.58602/jaiti.v4i2.271 Optimasi Metode Klasifikasi dengan Particle Swarm Optimization (PSO) dan Perbandingan Split Data untuk Prediksi Penyakit Diabetes https://ejournal.techcart-press.com/index.php/jaiti/article/view/266 <p>Diabetes merupakan suatu penyakit menahun (kronis) berupa gangguan metabolis yang ditandai dengan kadar gula darah yang melebihi batas normal. Prediksi awal penyakit Diabetes menjadi sesuatu yang penting agar tidak menjadi ancaman serius bagi penderitanya. Penelitian ini akan melakukan optimasi metode Machine Learning dengan reduksi fitur Particle Swarm Optimazation (PSO) dan perbandingan dataset training dan dataset testing untuk menemukan metode dengan performance terbaik dalam prediksi penyakit Diabetes. Metode Machine Learning yang dimaksud adalah metode klasifikasi Decision Tree (DT), Iterative Dichotomiser 3 (ID3), Random Forest (RF), Support Vector System (SVM), Naïve Bayes (NB), K-Nearest Neighbor (KNN) dan Neural Network (NN). Komparasi dilakukan dengan cara pada tiap matode, dataset yang digunakan dibagi menjadi data training dan data testing (validasi) dengan perbandingan 70:30, 80:20 dan 90:10. Pengukuran dengan cara tiap perbandingan dilihat nilai akurasi dan nilai AUC (kurva ROC) tertinggi pada data testing untuk menemukan performance terbaik. Dari olah data yang dilakukan diketahui untuk nilai Akurasi perbandingan data 70:30 metode DT, ID3 dan RF yang menghasilkan nilai tertinggi sama yaitu 96,79%. Untuk perbandingan 80:20 metode DT dan ID3 menghasilkan nilai tertinggi yaitu 97,12%. Dan perbandingan 90:10, metode ID3 menghasilkan nilai tertinggi yaitu 96,19%. Untuk nilai AUC metode RF menghasilkan nilai tertinggi untuk 3 perbandingan yaitu 0,995; 0,995 dan 0,998. Selanjutnya diketahui bahwa metode RF dan NN menjadi metode yang paling stabil dalam pengolahan menggunakan split data 70:30, 80:20 dan 90:10 ditinjau dari nilai akurasi dan nilai AUC.</p> Elly Muningsih Sutrisno Sutrisno Vembria Rose Handayani Cindyra Galuhwardani Copyright (c) 2026 Elly Muningsih, Sutrisno Sutrisno, Vembria Rose Handayani, Cindyra Galuhwardani https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 259 268 10.58602/jaiti.v4i2.266 A Pythagorean Fuzzy-Based MUNRA Method for Handling Uncertainty in Complex Decision Environments https://ejournal.techcart-press.com/index.php/jaiti/article/view/273 <p>This research develops the Pythagorean Fuzzy Multi-Normalized Rating Analysis (PF-MUNRA) method as a novel approach to address uncertainty and ambiguity in multi-criteria decision making. The main contribution of this study lies in the integration of Pythagorean Fuzzy Sets with a multi-normalization framework consisting of linear, vector, and non-linear normalization within a single decision-making model, enabling more flexible, comprehensive, and unbiased evaluation results compared to conventional single-normalization approaches. This method integrates the concept of Pythagorean Fuzzy Sets, which can represent degrees of membership and non-membership more flexibly, with the multi-normalization approach in MUNRA. Unlike previous studies that generally apply fuzzy environments and normalization techniques separately, the proposed PF-MUNRA simultaneously combines fuzzy uncertainty handling, multi-normalization mechanisms, and objective weighting to improve ranking consistency and decision robustness. In addition, weighted aggregation is used to produce more accurate preference values and reflect the relative importance of each criterion. The experimental results demonstrate that PF-MUNRA produces stable alternative rankings with Spearman correlation values ranging from 0.9464 to 1.0000 under various weight-change scenarios, indicating a very strong level of ranking consistency and robustness. Comparative analysis shows changes in alternative positions that reflect the capability of the proposed method to capture data complexity more effectively than the initial approach, while sensitivity analysis confirms that variations in criterion weights do not significantly affect the final ranking results, thereby proving that PF-MUNRA has high stability and reliability in dynamic and uncertain decision-making environments.</p> Setiawansyah Setiawansyah Copyright (c) 2026 Setiawansyah Setiawansyah https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 269 286 10.58602/jaiti.v4i2.273 Klasifikasi Serangan Web Berdasarkan Log Web Application Firewall (WAF) Menggunakan Support Vector Machine (SVM) https://ejournal.techcart-press.com/index.php/jaiti/article/view/262 <p>Pertumbuhan aplikasi <em>web</em> yang semakin pesat turut meningkatkan risiko ancaman keamanan siber terhadap layanan berbasis internet. Berbagai jenis serangan seperti <em>SQL Injection (SQLi)</em> dan <em>Cross-Site Scripting (XSS) </em>masih menjadi ancaman utama yang dapat mengganggu keamanan maupun ketersediaan sistem <em>web</em>. Penelitian ini bertujuan menerapkan pendekatan <em>machine learning</em> untuk melakukan klasifikasi serangan <em>web</em> menggunakan data <em>log</em> dari <em>Web Application Firewall</em> <em>(WAF) </em>berbasis <em>ModSecurity</em>. Data penelitian diperoleh dari audit <em>log</em> <em>ModSecurity</em> dalam format JSON yang berisi aktivitas request dan response pada <em>web</em> server. Tahapan penelitian meliputi pengumpulan data, <em>preprocessing</em>, ekstraksi fitur, labeling, <em>feature</em> <em>engineering</em> menggunakan <em>TF-IDF</em>, pembagian dataset dengan train-test split, pemodelan menggunakan algoritma <em>Support Vector Machine</em> <em>(SVM)</em>, serta evaluasi performa model menggunakan confusion matrix, accuracy, precision, <em>recall</em>, dan <em>F1-score</em>. Dataset yang digunakan terdiri atas 3467 data <em>traffic web</em> dengan kategori <em>SQL Injection</em>, <em>XSS</em>, dan Normal. Berdasarkan hasil pengujian, model SVM mampu menghasilkan tingkat akurasi sebesar 94,52% dalam proses klasifikasi <em>traffic web</em>. Model menunjukkan performa sangat baik pada pendeteksian serangan <em>SQL Injection</em> dengan <em>recall</em> sebesar 1,00 dan nilai <em>F1-score</em> sebesar 0,97. Akan tetapi, performa pada kategori Normal masih relatif rendah karena distribusi data yang tidak seimbang. Hasil penelitian menunjukkan bahwa analisis <em>log</em> <em>ModSecurity</em> yang dipadukan dengan <em>machine learning</em> dapat dimanfaatkan sebagai pendekatan alternatif untuk mendukung deteksi serangan <em>web</em> secara otomatis.</p> Nuroji Nuroji Tirta Anhari Copyright (c) 2026 Nuroji Nuroji, Tirta Anhari https://creativecommons.org/licenses/by-sa/4.0 2026-06-21 2026-06-21 4 2 287 299 10.58602/jaiti.v4i2.262