Search Articles & Publications

Showing 219 articles found for "Specific"

OPTIMIZING RETRIEVAL-AUGMENTED GENERATION FOR DOMAIN-SPECIFIC KNOWLEDGE SYSTEMS THROUGH FINE-TUNING AND PROMPT ENGINEERING

Ahmad Fajri, Rila Mandala
Abstract: Abstract: This study discusses the optimization of RAG for a FAQ system in the field of information technology product security certification at BSSN. Although LLM generate reliable responses, they often lack up-to-date… and domain-specific knowledge, which can be addressed through the RAG approach. This research aims to optimize a domain-specific RAG system by improving embedding performance, enhancing prompt robustness, and increasing retrieval accuracy. The research methods consist of three stages. The first stage involves fine-tuning the bge-m3 embedding model and evaluating its performance using MRR, Recall, and AUC. The second stage applies prompt engineering techniques, namely the SRSM and Autodefense, to mitigate direct-injection and escape-character prompt injection attacks. The third stage evaluates the proposed RAG system using Precision, Recall, and F1-Score metrics against four baseline models. The results of research show that the fine-tuned embedding model achieves higher performance than the original model, with MRR@1 and Recall@1 values of 0.80 and an AUC@100 of 0.7023. In addition, the proposed prompt engineering techniques demonstrate robustness against prompt injection attacks, while the overall RAG system attains a perfect Precision, Recall, and F1-Score of 1.00. In conclusion, the proposed approach effectively enhances retrieval accuracy, embedding quality, and system security, resulting in a more reliable RAG-based FAQ system for information technology product security certification. Keywords: embedding fine-tuning; large language model; prompt engineering; prompt injection mitigation; retrieval-augmented generation   Abstrak: Studi ini membahas optimasi RAG untuk sistem FAQ di bidang sertifikasi keamanan produk teknologi informasi di BSSN. Meskipun LLM menghasilkan respons yang andal, mereka seringkali kurang memiliki pengetahuan terkini dan spesifik domain, yang dapat diatasi melalui pendekatan RAG. Penelitian ini bertujuan untuk mengoptimalkan sistem RAG spesifik domain dengan meningkatkan kinerja embedding, meningkatkan ketahanan prompt dan meningkatkan akurasi pengambilan. Metode penelitian terdiri dari tiga tahap. Tahap pertama melibatkan fine-tuning model embedding bge-m3 dan mengevaluasi kinerjanya menggunakan Mean Reciprocal Rank (MRR), Recall, dan AUC. Tahap kedua menerapkan teknik rekayasa prompt, yaitu Self- SRSM dan Autodefense, untuk mengurangi serangan direct-injection dan escape-character prompt injection. Tahap ketiga mengevaluasi sistem RAG yang diusulkan menggunakan metrik Presisi, Recall, dan F1-Score terhadap empat model dasar. Hasil penelitian menunjukkan bahwa model embedding yang disempurnakan mencapai kinerja yang lebih tinggi daripada model asli, dengan nilai MRR@1 dan Recall@1 sebesar 0,80 dan AUC@100 sebesar 0,7023. Selain itu, teknik rekayasa prompt yang diusulkan menunjukkan ketahanan terhadap serangan injeksi prompt, sementara sistem RAG secara keseluruhan mencapai Presisi, Recall, dan F1-Score sempurna sebesar 1,00. Kesimpulannya, pendekatan yang diusulkan secara efektif meningkatkan akurasi pengambilan, kualitas embedding dan keamanan sistem, menghasilkan sistem FAQ berbasis RAG yang lebih andal untuk sertifikasi keamanan produk teknologi informasi. Kata kunci: penyempurnaan embedding; model bahasa besar; rekayasa prompt; mitigasi injeksi prompt; retrieval-augmented generation

A COMPARATIVE ANALYSIS OF OPTIMIZED NEURAL NETWORK AND LARGE-SCALE LANGUAGE MODELS FOR MUSIC GENRE CLASSIFICATION

Marzuqi, Ahmad Naufal Luthfan, Nastiti , Vinna Rahmayanti Setyaning
Abstract: Abstract: The rapid growth of the digital music industry requires accurate music genre classification systems to enhance user experience in streaming services. This study compares a domain-specific Long Short-Term Memory… (LSTM) network with three Large Language Models (LLMs)—HuBERT, WavLM, and WAV2Vec 2.0—for Music Genre Classification (MGC). The LSTM model was trained using Mel-spectrograms transformed from the GTZAN dataset, while the LLMs were fine-tuned using a smaller set of raw audio samples due to computational constraints. All models were tested on datasets with identical genre labels to ensure a fair evaluation. Results show that the LSTM model achieved the highest accuracy of 97.10%, outperforming HuBERT (86.00%), WavLM (83.00%), and WAV2Vec 2.0 (80.00%). The LSTM demonstrated superior generalization and stability without overfitting, while the LLMs struggled to differentiate between genres with similar acoustic characteristics. These findings indicate that general-purpose pre-trained models, although powerful, are less effective in music-specific tasks due to domain mismatch. Therefore, incorporating music-specific features and architectures remains essential for achieving higher accuracy and reliability in automatic genre classification systems. Keywords: audio large language models; comparative deep learning; music genre classification.   Abstrak: Pertumbuhan industri musik digital yang pesat menuntut sistem klasifikasi genre musik yang akurat untuk meningkatkan pengalaman pengguna dalam layanan streaming. Penelitian ini dilatarbelakangi oleh perkembangan pesat model pembelajaran mendalam, khususnya jaringan LSTM dan model bahasa berskala besar LLM seperti HuBERT, WavLM, dan WAV2Vec 2.0, yang telah menunjukkan kemampuan representasi audio yang kuat. Tujuan penelitian ini ini membandingkan jaringan Long Short-Term Memory (LSTM) khusus domain dengan tiga model Large Language Models (LLM)—HuBERT, WavLM, dan WAV2Vec 2.0—untuk tugas Klasifikasi Genre Musik (MGC). Metode penelitian melibatkan pelatihan LSTM menggunakan data Mel-spectrogram hasil transformasi dari dataset GTZAN, sementara LLM disesuaikan (fine-tuning) menggunakan data audio mentah dalam jumlah lebih kecil karena keterbatasan komputasi. Seluruh model diuji pada dataset dengan label genre yang sama untuk memastikan evaluasi yang adil. Hasil penelitian menunjukkan bahwa model LSTM mencapai akurasi tertinggi sebesar 97,10%, sedangkan model HuBERT, WavLM, dan WAV2Vec 2.0 masing-masing memperoleh 86,00%, 83,00%, dan 80,00%. Model LSTM menunjukkan kemampuan generalisasi yang lebih baik tanpa overfitting, sedangkan model LLM cenderung kesulitan membedakan genre dengan karakteristik akustik yang mirip. Kesimpulan penelitian ini adalah ketidaksesuaian domain secara signifikan membatasi performa model umum saat diterapkan pada tugas berbasis musik. Oleh karena itu, penggunaan fitur dan arsitektur khusus musik sangat penting dalam membangun sistem klasifikasi genre yang lebih akurat. Kata kunci: klasifikasi genre musik; model bahasa besar; perbandingan pembelajaran mendalam.

USE OF TASK-CENTERED SYSTEM DESIGN IN THE INTERFACE DESIGN OF THE POPULATION DEMOGRAPHIC DATA INFORMATION SYSTEM

Muhammad Azmi Zaky, Allsela Meiriza, Dinda Lestarini, Pacu Putra, Nabila Rizki Oktadini
Abstract: Abstract: The rapid development of information and communication technology has prompted the government to provide digital-based services, including in the management of demographic data. This study aims to apply the Task-Centered… k-Centered System Design (TCSD) method in designing the Muara Enim Regency Demographic Data Information System. The TCSD method was chosen to ensure that the prototype design process was systematic and focused on user needs and tasks. The research stages included identification, user-centered needs analysis, scenario-based design, and walkthrough evaluation. The designed prototype supports several main tasks, including viewing demographic statistics, searching for specific data, submitting data download requests, and contacting the admin. The evaluation was conducted through online usability testing using the Maze platform with the System Usability Scale (SUS) instrument involving 13 respondents. The evaluation results showed an average SUS score of 78.5, which falls into the “good” category. This confirms that the interface design has met usability standards, is user-friendly, and is capable of supporting user needs in accessing and managing demographic data. Thus, the application of the TCSD method has proven to be effective in producing an interface design that is focused on user tasks and can be the basis for further system development. Keywords: system usability scale; task centered system design; user interface

IMPLEMENTATION OF RANDOM FOREST CLASSIFIER FOR STUDENT GRADUATION CLASSIFICATION

Zaidan Putra, Bazil, Nur Fajri, Ika, Nugroho, Agung
Abstract: Abstract: Higher education plays an essential role in improving human resource quality, one of which is through the institution’s ability to monitor and predict student graduation outcomes. This study does not focus on a… a specific university but utilizes the publicly available Students Performance in Exams dataset from Kaggle, consisting of 1,000 student records containing mathematics, reading, and writing scores, along with demographic attributes such as gender, parental education level, lunch type, and test preparation participation. The data were processed through a feature engineering stage by adding an average score variable as an early indicator of graduation status. A predictive model was developed using the Random Forest Classifier, achieving an accuracy of 94.5%. The final model was integrated into a Streamlit-based web application to provide an accessible tool for academic stakeholders. The results indicate that the proposed model can serve as an effective decision-support tool for early evaluation of students’ likelihood of graduation. Keywords: prediction; random forest classifier, streamlit, student graduation.     Abstrak: Pendidikan tinggi memegang peran penting dalam peningkatan kualitas sumber daya manusia, salah satunya melalui kemampuan institusi dalam memantau dan memprediksi tingkat kelulusan mahasiswa. Penelitian ini tidak berfokus pada perguruan tinggi tertentu, melainkan menggunakan dataset publik Students Performance in Exams dari Kaggle yang berisi 1.000 data mahasiswa, terdiri atas nilai matematika, membaca, menulis, serta atribut demografis seperti gender, tingkat pendidikan orang tua, jenis makan siang, dan partisipasi kursus persiapan. Data diolah melalui tahap feature engineering dengan menambahkan variabel average score sebagai indikator awal kelulusan. Model prediksi dibangun menggunakan algoritma Random Forest Classifier, yang menghasilkan tingkat akurasi sebesar 94,5%. Model ini kemudian diimplementasikan ke dalam aplikasi web berbasis Streamlit untuk memberikan layanan prediksi yang mudah diakses oleh pihak akademik. Hasil penelitian menunjukkan bahwa model mampu digunakan sebagai alat pendukung keputusan untuk melakukan evaluasi dini terhadap potensi kelulusan mahasiswa.   Kata kunci: kelulusan mahasiswa; prediksi; random forest classifier; streamlit.

MACHINE LEARNING CONTENT-BASED FILTERING WOMEN EMPOWERING RECOMMENDATIONS ON YOUTUBE

Yuliana, Yuliana, Mira, Mira, Hari Kristianto, Aloysius
Abstract: Abstract: YouTube is one of the most popular video streaming platforms, but it has constraints that can cause problems when clients have difficulty finding content according to their wishes. The main objective of this study… udy is to increase user capacity in viewing content specifically in the field of women's empowerment. By using content-based filtering techniques, the system will analyze user preferences and interests through recommendations for women's empowerment content. The data source is via the YouTube API and is analyzed using PHP programming content-based filtering techniques. The system's recommendations provide a list of women's empowerment content with a user request display. The results of the research evaluation obtained a precision value of 62%, meaning that the recommendations match the topic being searched for, namely women's empowerment. The recall value of 84% indicates that the system has succeeded in finding relations from the database. The f1-score value of 72% indicates that there is a balance between precision and recall, meaning that a system is needed that is not only accurate but also complete. While the cosine value shows a score of 0.7071 approaching the maximum value (1.0). The recommendation of the content-based filtering method produces quite effective women's empowerment content. Keywords: content-based filtering, recommendations, women Empowerment, youtube  

ANALYSIS OF NEURAL NETWORK ALGORITHM IN URBAN AIR QUALITY PREDICTION

Anggraeni, Dewi, Azmi, Sri Rezki Maulina
Abstract: Abstract: Air quality in urban areas is becoming an increasingly important issue considering its impact on human health and the environment. The rapid increase in air pollution requires effective methods to predict air quality… uality in order to take appropriate mitigation measures. This study aims to analyze the use of Neural Network (NN) algorithms in predicting air quality in cities. The method used is the application of the NN model, especially the Multilayer Perceptron (MLP), which is trained using historical air quality data such as dust particle levels (PM10, PM2.5), carbon monoxide (CO) gas, and temperature. The data used in this study came from urban air quality monitoring stations collected over a period of time. The results show that the Neural Network algorithm can provide quite accurate predictions of air quality with a low Mean Absolute Error (MAE) value, showing the effectiveness of the model in predicting f fluctuations in air quality. The conclusion of this study is that Neural Network algorithms, specifically MLPs, are an effective tool for air quality prediction, which can be used as a basis for urban air quality management policies.  Keywords: air quality;  neural network; prediction; multilayer perceptron (MLP)    Abstrak: Kualitas udara di perkotaan menjadi isu yang semakin penting mengingat dampaknya terhadap kesehatan manusia dan lingkungan. Peningkatan polusi udara yang pesat memerlukan metode yang efektif untuk memprediksi kualitas udara guna mengambil langkah mitigasi yang tepat. Penelitian ini bertujuan untuk menganalisis penggunaan algoritma Neural Network (NN) dalam memprediksi kualitas udara di perkotaan. Metode yang digunakan adalah penerapan model NN, khususnya Multilayer Perceptron (MLP), yang dilatih menggunakan data kualitas udara historis seperti kadar partikel debu (PM10, PM2.5), gas karbon monoksida (CO), dan suhu. Data yang digunakan dalam penelitian ini berasal dari stasiun pemantauan kualitas udara di perkotaan yang dikumpulkan selama periode waktu tertentu. Hasil penelitian menunjukkan bahwa algoritma Neural Network dapat memberikan prediksi yang cukup akurat terhadap kualitas udara dengan nilai Mean Absolute Error (MAE) yang rendah, menunjukkan efektivitas model dalam memprediksi fluktuasi kualitas udara. Simpulan dari penelitian ini adalah bahwa algoritma Neural Network, khususnya MLP, merupakan alat yang efektif untuk prediksi kualitas udara, yang dapat digunakan sebagai dasar untuk kebijakan pengelolaan kualitas udara di perkotaan Kata kunci: kualitas udara; neural network; prediksi; multilayer perceptron (MLP)

MOBILE APPLICATION DESIGN TO PROMOTE CULTURAL TOURISM IN WEST SUMBA

Rangga, Andreas Ariyanto, Dapadeda, Ardiyanto
Abstract: Abstract: West Sumba Regency has many traditional villages, local dances and bu-daya festivals that can be seen as cultural tourist attractions. This is indicated by the increasing number of people visiting cultural tourist… ist attractions in West Sumba. However, there are still few promotional and information services available for tourists related to cultural tourism to visit, such as location information, events and access to cultural tourism spots in West Sumba. Therefore, the purpose of this research is to create a mobile phone-based application that can promote bu-culture tourism in West Sumba Regency. The specific objective is to create a platform that makes it easy for tourists to explore and understand the cultural richness and cultural tourism destinations of West Sumba. This research uses a user-centered design (UCD) approach to create a successful mobile application. This application will help tourists get information about cultural tourism destinations in West Sumba. In addition, the app has additional features, such as maps, traditional villages, local dances, cultural festivals, event schedules and travel agents.   Keywords: Cultural Tourism; Design; Mobile Apps; west sumba   Abstrak: Kabupaten Sumba Barat memiliki banyak kampung adat, tarian lokal, dan festival budaya yang dapat dilihat sebagai tempat wisata budaya. Ini ditunjukkan oleh peningkatan jumlah orang yang mengunjungi tempat-tempat wisata budaya di Sumba Barat. Namun, masih sedikit layanan promosi dan informasi yang tersedia bagi para wisatawan terkait wisata budaya yang ingin dikunjungi, seperti informasi lokasi, acara dan akses ke tempat-tempat wisata budaya di Sumba Barat. Oleh karena itu, tujuan dari penelitian ini adalah untuk membuat aplikasi berbasis ponsel yang dapat mempromosikan wisata budaya di Kabupaten Sumba Barat. Tujuan khusus adalah untuk membuat sebuah platform yang memudahkan wisatawan untuk mengeksplorasi dan memahami kekayaan budaya dan destinasi wisata budaya Sumba Barat. Penelitian ini menggunakan pendekatan desain berbasis pengguna (UCD) untuk membuat aplikasi mobile yang berhasil. Aplikasi ini akan membantu wisatawan mendapatkan informasi tentang destinasi wisata budaya di Sumba Barat. Selain itu, aplikasi ini memiliki fitur tambahan, seperti peta, kampung adat, tarian lokal, festival budaya, jadwal acara, dan agen perjalanan wisata.   Kata kunci: aplikasi seluler; desain; sumba barat; wisata budaya  

OPTIMIZATION OF CART ALGORITHM BASED ON ANT BE COLONY FEATURE SELECTION FOR STUNTING DIAGNOSIS

Subarkah, Pungkas, Ikhsan, Ali Nur, Wahyudi, Rizki, Rofiqoh, Dayana
Abstract: Abstract: One of the main health problems in children is stunting which is one of the concerns in the Sustainable Development Goals (SDGs). Specifically in Indonesia, the prevalence of stunting in 2024 is 21.6%. This figure… ure is still relatively high, because the target prevalence of stunting is 14%. This study aims to implement machine learning knowledge through the Classification And Regression Trees (CART) algorithm based on Ant Be Colony (ABC) feature selection which aims to determine the increase in accuracy in analyzing stunting datasets. The data used comes from Kaggle which consists of 16500 datasets. The dataset consists of gender, age, birth length, birth weight, body length, body weight, breastfeeding and stunting status. The research methods used are data collection, data preprocessing, classification, and evaluation using K-fold cross validation. The results obtained in this research are the implementation of the CART algorithm obtained a value of 89.86% and the results of CART with Ant Be Colony (ABC) feature selection, which obtained an accuracy value of 93.65%. This shows that there is an increase in the accuracy value in the use of CART algorithm optimization and Ant Be Colony (ABC) feature selection by 3.76%. With the research results that have been obtained, it can be categorized as excellent accuracy value excellent. It is hoped that further research can be carried out by adding other classification algorithms or adding feature selection.             Keywords: classification; feature selection; optimazation; stunting   Abstrak: Salah satu masalah kesehatan utama pada anak adalah stunting yang menjadi salah satu perhatian dalam Sustainable Development Goals (SDGs). Khusus di Indonesia angka Pravelensi stunting pada tahun 2024 di angka 21.6%. Angka ini masih tergolong tinggi, karena target angka pravelensi stunting ialah 14%. Penelitian ini bertujuan untuk mengimplementasikan pengetahuan machine learning melalui algoritma Classification And Regression Trees (CART) berbasis seleksi fitur Ant Be Colony (ABC) yang bertujuan untuk mengetahui peningkatan akurasi dalam menganalisis dataset stunting. Data yang digunakan bersumber dari Kaggle yang terdiri dari 16500 dataset. Dataset terdiri dari jenis kelamin, usia, panjang lahir, berat lahir, panjangg badan, berat badan, menyusui dan status stunting.  Metode penelitian yang digunakan adalah pengumpulan data, preprocessing data, klasifikasi, dan evaluasi menggunakan K-fold cross validation. Hasil yang diperoleh pada penelitian ini adalah Implementasi algoritma CART memperoleh nilai sebesar 89,86% dan hasil seleksi fitur CART dengan Ant Be Colony (ABC) memperoleh nilai akurasi sebesar 93,65%. Hal ini menunjukkan adanya peningkatan nilai akurasi pada penggunaan optimasi algoritma CART dan pemilihan fitur Ant Be Colony (ABC) sebesar 3,76%. Dengan hasil penelitian yang telah diperoleh dapat dikategorikan nilai akurasi yang diperoleh sangat baik. Diharapkan dapat dilakukan penelitian selanjutnya dengan menambahkan algoritma klasifikasi lain atau menambahkan seleksi fitur.   Kata kunci: klasifikasi; optimalisasi; seleksi fitur; stunting

COMPARATIVE ANALYSIS OF K-MEANS, X-MEANS AND K-MEDOIDS IN CLASSIFYING MARRIAGE CHOICED ADMIST QUARTER-LIFE CRISIS

Ariza, Disya Nurul, Ningsih, Rahayu, Muryani, Sri, Ferliyanti, Herlina, Wahidin, Ahmad Jurnaidi
Abstract: Abstract: Bekasi Regency, being one of the key cities in Indonesia, offers a suitable setting to study the intricacies of marriage decision-making during a quarter-life crisis. This study focuses on the application of clustering… ustering algorithms to categorize individuals based on their marriage choices. Data was collected from a questionnaire completed by 110 respondents from Bekasi Regency, specifically individuals aged 18 to 30 who are single, including 80 women and 30 men. Data analysis was conducted using the RapidMiner software to evaluate the effectiveness of three clustering algorithms K-Means, X-Means, and K-Medoids in categorizing marriage decision patterns among young people experiencing a Quarter Life Crisis in Bekasi Regency. Results indicate that each algorithm has its own strengths and limitations in handling Quarter Life Crisis data.The results of the analysis show that the K-medoids algorithm provides the best clustering results with the lowest DBI value of 0.195, followed by the X-Means algorithm with a value of 0.199 and K-Means with a value of 0.207. These results can help understand the pattern of marriage decisions in the Quarter Life Crisis phase and help provide insights for policymakers in Bekasi Regency to make more effective intervention programs.             Keywords: K-Means; K-Medoids; X-Means     Abstrak: Sebagai salah satu kota besar di Indonesia, Kabupaten Bekasi memberikan konteks yang tepat untuk mempelajari kompleksitas pengambilan keputusan pernikahan di tengah krisis seperempat usia. Penelitian ini berfokus pada pemanfaatan algoritma clustering untuk mengelompokkan individu berdasarkan pilihan pernikahan mereka. Data diambil dari kuesioner yang diisi oleh 110 responden di Kabupaten Bekasi, yang terdiri dari individu lajang berusia 18 hingga 30 tahun, yaitu 80 perempuan dan 30 laki-laki. Analisis data dilakukan dengan perangkat lunak RapidMiner untuk mengevaluasi efektivitas tiga algoritma pengelompokan—K-Means, X-Means, dan K-Medoids—dalam mengelompokkan pola keputusan pernikahan di kalangan pemuda yang menghadapi Quarter Life Crisis di Kabupaten Bekasi. Hasilnya menunjukkan bahwa setiap algoritma memiliki keunggulan dan kelemahannya masing-masing dalam memproses data Quarter Life Crisis. Hasil analisis menunjukkan bahwa algoritma K-medoids memberikan hasil clustering terbaik dengan nilai DBI terendah yaitu 0.195, diikuti oleh algoritma X-Means dengan nilai 0.199 dan K-Means dengan nilai 0.207. Hasil ini dapat membantu memahami pola keputusan menikah pada fase Quarter Life Crisis dan membantu memberikan wawasan bagi pembuat kebijakan di Kabupaten Bekasi membuat program intervensi yang lebih efektif.     Kata kunci: K-Means; K-Medoids; X-Means  

ANALYSIS OF DC MOTOR ROTATION SPEED ON THE BURNER STOVE FLAME USING OIL AS FUEL: A CASE STUDY IN TOMUANHOLLBUNG VILLAGE

Afandi, Adi Mas, Ananda, Ricki
Abstract: Abstract: Due to the successful conversion from LPG stoves to electric stoves, which are unsupported by the village’s electricity supply—averaging 450 watts—while government-provided stoves have a power rating of 900… wer rating of 900 watts, making them unsuitable for use in Tomuanhollbung Village. To address this issue, the research team designed a stove that uses waste oil (35,000 kJ/kg) and incorporated embedded system technology, utilizing the French design research method. The research results indicated that by using a DC motor with specifications of 12VDC/3A, 3800 rpm, and a valve diameter of 9.7 x 9.5 x 3.3 cm, a power output of 36W was achieved, with a pressure of 288.7 Pa or 0.00289 bar and an angular velocity of 398.1 rad/s. When the burner stove was first ignited, the flame appeared yellow. With a voltage of 3.7VDC/3A, it produced a rotation of 3795 rpm, a yellow flame, a flame height of 8 cm from the stove, and a flame width diameter of approximately 18 cm. At a voltage of 3.7VDC, the pressure measured 0.000964 bar, with a flame temperature of 275°C (as measured by an infrared sensor). A voltage of 7.4VDC resulted in a pressure of 0.0192 bar and a flame temperature of 350°C, while a voltage of 11.8VDC produced a pressure of 0.0289 bar with a maximum flame temperature of 420°C. For fuel consumption over 10 minutes at 3.7VDC/3A, the flame height reached 7-10 cm, with a wind pressure of 0.000964 bar and a fuel consumption of 0.237 grams.                                                           Keywords : embedded system burner stove; microcontroller; snail dc  motor.   Abstrak: Dilatar belakangi masalah kelangkaan elpiji didesa tomuanhollbung, serta tidak berhasilnya konversi kompor elpiji  menuju kompor listrik karena daya listrik rumah masyarakat berdaya 450wat,  daya kompor pemerintah 900watt, sehingga tidak bisa digunakan. Dari masalah tersebut, tujuan utama penelitian ini merancang kompor berbahan bakar oli bekas, yang mampu mengatasi masalah kelangkaan gas elpiji dan tidak terpakainya kompor dari pemerintah psat.  Metode penelitian ini menggnakan metode prototype sehingga menghasilkan satu produk kompor burner dengan mengatur kecepatan motor DC. Hasil penelitian mendapati data pengujian mendapati bahwa dengan menggunakan motor DC Keong spesifikasi tegangan 12VDC/3A, 3800 rpm, dengan diameter katup 9,7 x 9.5 x 3.3cm menghasilkan daya 36W, dengan tekanan 288.7Pa atau 0.00289bar kecepatan sudut 398.1 rad/s. Ketika kompor burner pertama kali dinyalakan, warna api berwarna kunig. Jika tegangan 3.7VDC/3A akan menghasilkan putaran 3795rpm  nyala api kuning, tinggi api 8cm dari tunggku kompor, dan diameter lebar keluarnya api dari tungku berkisar 18cm. Tegangan 3,7VDC maka menghasilkan tekanan 0.000964bar, nyala api 275 0C (pembacaan sensor infrared).  Tegangan 7.4VDC menghasilkan tekanan 0.0192 bar, nyala api 350 0C, dan Tegangan 11.8 VDC tekanan 0.0289 bar, nyala api maksimal 420 0C.  Untuk konsumsi bahan bakar selama 10 menit, dengan tegangan 3.7VDC/3A mendapati nyala api setinggi 7-10 cm, tekanan angin senilai 000964 bar dan konsumsi bahan bakar 0.237 gram. Kesimpulan penelitian ini mendapati bahwa nyala api kompor burner bisa di atur seperti nyala api kompor elpiji pada umum nya.                                                      Kata kunci: kompor burner embedded system, microcontroller, motor dc keong.