AGENTS: Journal of Artificial Intelligence and Data Science

IMPLEMENTASI SISTEM PANDU (PELAYANAN ADUAN DIGITAL) DENGAN RESPON OTOMATIS BERBASIS AI UNTUK MENINGKATKAN KUALITAS LAYANAN PUBLIK PADA KELURAHAN KEBONWARU

shafa salsabila

Public complaint services in Kebonwaru Village still face several challenges, such as slow response times, limited staff capacity in handling multiple complaints simultaneously, and the absence of structured documentation, which impacts the quality of public services. This study aims to implement the PANDU System (Digital Complaint Service) based on artificial intelligence to automate responses and organize complaint management. The research method employed is descriptive through observation, interviews, and literature study, while system development follows the Waterfall model consisting of requirements analysis, design, implementation, and testing. The system is designed with WhatsApp integration and Gemini AI using the RAG (Retrieval Augmented Generation) approach to generate relevant automatic responses and provide an escalation feature to staff when necessary. Testing was conducted using Black-Box Testing to assess functionality and Top-1 Accuracy to evaluate AI response accuracy. The findings show that the system can automate responses, improve response accuracy, document complaints in a structured manner, and support follow-up escalation. Thus, the implementation of the PANDU System has proven to enhance the quality of public services, strengthen public trust, and foster complaint services that are more responsive and citizen-oriented.

KLASIFIKASI CITRA WAJAH MAHASISWA STMIK “AMIKBANDUNG” MENGGUNAKAN MOBILENETV2

Rinrin Sri Maulida

Presensi mahasiswa merupakan bagian penting dalam mendukung kelancaran proses pembelajaran. Di STMIK AMIK Bandung, proses pencatatan presensi masih dilakukan secara manual sehingga memakan waktu, berisiko menimbulkan kesalahan, serta membuka peluang terjadinya kecurangan. Penelitian ini mengusulkan penerapan teknologi pengenalan wajah berbasis model MobileNetV2 untuk mencatat presensi secara otomatis, cepat, dan akurat. Dataset wajah mahasiswa dikumpulkan secara langsung dengan variasi sudut pandang dan pencahayaan. Tahapan penelitian meliputi deteksi dan pemotongan wajah menggunakan MTCNN, ekstraksi ciri wajah dengan MobileNetV2, serta klasifikasi identitas melalui Logistic Regression. Data dibagi menjadi enam puluh empat persen untuk pelatihan, enam belas persen untuk validasi, dan dua puluh persen untuk pengujian. Evaluasi kinerja dilakukan menggunakan metrik akurasi, precision, recall, dan f1-score. Hasil pengujian terhadap seratus empat puluh enam gambar uji menunjukkan akurasi sebesar delapan puluh satu koma lima satu persen dengan nilai rata-rata precision, recall, dan F1-score di atas nol koma delapan lima. Dengan demikian, penelitian ini membuktikan bahwa MobileNetV2 dapat diimplementasikan sebagai solusi efektif untuk sistem presensi mahasiswa berbasis pengenalan wajah.

English

Reza Fitriansyah Reza

[PREDICTION OF SLEEP DISORDER RISK FACTORS USING MACHINE LEARNING APPROACH: LOGISTIC REGRESSION AND GRADIENT BOOSTING] Sleep disorders are one of the major public health issues with broad implications for quality of life, productivity, and chronic disease risks. This study aims to predict risk factors of sleep disorders using a machine learning approach with survey data from the National Sleep Foundation. The research process involved data Cleaning, traNSFormation, normalization, and splitting into training (80%) and testing (20%) sets. Two algorithms were applied, Logistic Regression and Gradient Boosting, and their performance was evaluated using Accuracy, Precision, Recall, and F1-score metrics. SHAP analysis was also employed to assess variable contributions to model predictions. The results indicate that Gradient Boosting outperformed Logistic Regression, achieving perfect Accuracy and F1-score (1.00), while Logistic Regression only reached 0.70. SHAP analysis revealed that sleep duration and quality are the most influential factors, followed by caffeine consumption and age. Therefore, Gradient Boosting not only provides accurate classification but also comprehensive insights into key determinants of sleep disorders, serving as a foundation for more effective health interventions.

Prediksi Faktor Risiko Gangguan Tidur Menggunakan Pendekatan Machine Learning Logistic Regression dan Gradient Boosting

Reza Fitriansyah Reza, R Tommy Gumelar, Amrizal Amrizal — Fri, 29 Aug 2025 00:00:00 +0000

Sleep disorders are one of the major public health issues with broad implications for quality of life, productivity, and chronic disease risks. This study aims to predict risk factors of sleep disorders using a machine learning approach with survey data from the National Sleep Foundation. The research process involved data Cleaning, transformation, normalization, and splitting into training (80%) and testing (20%) sets. Two algorithms were applied, Logistic Regression and Gradient Boosting, and their performance was evaluated using Accuracy, Precision, Recall, and F1-score metrics. SHAP analysis was also employed to assess variable contributions to model predictions. The results indicate that Gradient Boosting outperformed Logistic Regression, achieving perfect Accuracy and F1-score (1.00), while Logistic Regression only reached 0.70. SHAP analysis revealed that sleep duration and quality are the most influential factors, followed by caffeine consumption and age. Therefore, Gradient Boosting not only provides accurate classification but also comprehensive insights into key determinants of sleep disorders, serving as a foundation for more effective health interventions.

Implementation of Fuzzy Tsukamoto in the Design of Nutrient Control and Monitoring System for Aquaponics Based on the Internet of Things

Abd Muqsith Hidayat, Faisal Akib , Faisal Faisal — Sat, 31 Aug 2024 00:00:00 +0000

Aquaponic is an important aquaculture technique because it is easy to apply, saves water, and allows the integration of plant roots to absorb waste nitrogen from fish waste as nutrients. However, temperature, pH, Total Dissolved Solids (TDS), and water level greatly affect plant growth. This research aims to design a control system to monitor plant nutrition and development in real-time using temperature, pH, TDS, and ultrasonic sensors and apply Tsukamoto Fuzzy model to overcome uncertainty in decision making based on sensor data. This research uses a quantitative approach with a design and development method. Data were collected through direct observation, interviews with aquaponic farmers, and related literature studies. The designed system successfully fulfills the need to control and monitor nutrients in aquaponic systems effectively. The system utilizes an ESP8266 module and various sensors (pH, TDS/PPM, temperature, and water level) to monitor water conditions in real-time and send the data to Firebase, which is then displayed on the application interface. Automatic control allows for quick adjustments to changing environmental conditions, ensuring an optimal environment for plant growth.