Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Donny Avianto has a total of 44 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Implementasi Convolutional Neural Network Untuk Pengenalan Tulisan Tangan Akasara Sunda Ngalangéna

    by Rahma Nur Azizah, Donny Avianto

    Efforts to preserve the Sundanese script as a cultural heritage face challenges in the digital era, one of which is the limited resources for pattern recognition. This research aims to develop an effective custom Convolutional Neural Network (CNN) model for the classification of handwritten Sundanese script. Facing the constraint of no available public dataset, this study utilizes a primary dataset (Swaraksara Dataset) created by the author, consisting of 6,500 handwritten images evenly distributed across 13 classes (combinations of the "Na" script with rarangkén). The methodology applied includes a comprehensive data preprocessing stage, covering grayscale conversion, resizing to 200x200 pixels, normalization, and data augmentation techniques to prevent overfitting. The custom CNN architecture was designed with five convolutional layers (filters 32 to 512) and the Adam optimizer. The experimental results show that the optimal configuration was achieved with a learning rate of 0.001 and 50 training epochs, resulting in very high model performance. In the evaluation using test data, the model achieved an accuracy of 99.54% with a loss value of 0.0175. The optimal performance of this model is driven by the quality of the primary dataset supported by comprehensive image preprocessing stages, thus ensuring clean, uniform, and significantly noise-free data input. Analysis of the confusion matrix and learning curves also confirmed the model's excellent generalization ability with no indications of overfitting. This model has been successfully implemented in the "Swaraksara" web application as a Sundanese script recognition system.

  • Open Access

    Speech-Based Virtual Assistant for Mental Health Support Through Natural Interaction

    by Dimas Rizqi Kurniawan, Donny Avianto

    Mental health is a significant global concern. Indonesia has reported high rates of depression and anxiety, compounded by limited emotional outlets. Although AI virtual assistants are prevalent in e-commerce and education, their application in mental health remains underexplored. Existing solutions are predominantly text-based and transactional, which restricts empathetic and natural interactions. This study develops a voice-based assistant by integrating Automatic Speech Recognition (ASR), a generative AI for empathetic responses, and a Text-to-Speech (TTS) module fine-tuned on an Indonesian dataset to adapt accent and prosody. The system underwent both technical evaluation and human testing to assess its feasibility and user experience. The results showed that the TTS model converged effectively with low loss. Human evaluation indicated 'good' interaction (MS = 3.91, SD = 0.02), 'good' AI responses (MS = 3.83, SD = 0.26), and 'fair' TTS naturalness (MOS = 3.27, SD = 0.05). Most participants found the assistant's responses meaningful, pleasant, and helpful in managing low to moderate anxiety. These results suggest that a voice-based assistant has the potential to support mental health in Indonesia. Future work should enhance speech naturalness and utilize a larger participant pool for evaluation.

  • Open Access

    ANALISIS KINERJA SUPPORT VECTOR MACHINE DAN NAÏVE BAYES CLASSIFIER DALAM KLASIFIKASI SENTIMEN DAN EMOSI PADA UMPAN BALIK MAHASISWA TERHADAP KINERJA DOSEN

    by Teguh Syafrudin, Arief Hermawan, Donny Avianto

    Pencatatan rekam medis yang efisien dan akurat sangat penting dalam menunjang layanan kesehatan yang berkualitas. Hal ini berlaku juga pada pelayanan kesehatan berbasis komunitas seperti pada klinik atau Praktik Bidan Mandiri (PBM). Tujuan penelitian ini adalah untuk mengembangkan sistem rekam medis elektronik berbasis web yang dapat menangani permasalahan yang biasa terjadi dalam sistem manual di PBM Leyna Merlina Sarkosi, Bandung. Pengembangan pembuatan aplikasi dilakukan menggunakan pendekatan waterfall, dimana tahapan yang dilakukan dimulai dari tahapan analisis kebutuhan, desain sistem, implementasi dengan React.js dan Laravel, serta diakhiri dengan pengujian fungsional terhadap seluruh fitur. Pengguna dari aplikasi ini ditujukan untuk dua tipe pengguna, yakni admin dan pasien. Sedangkan fitur yang dikelola dalam aplikasi ini meliputi pengelolaan data pasien, pencatatan layanan kesehatan, penjadwalan kunjungan, serta pembatasan akses informasi medis secara aman. Dari hasil pengujian yang sudah dilakukan menunjukkan seluruh fungsi dapat berjalan dengan baik, termasuk fitur kohort layanan dan akses data pasien berbasis NIK. Sistem juga mendapatkan respons positif dari pengguna, baik dari sisi kemudahan operasional maupun keamanan data. Penelitian ini memberikan kontribusi pada pengembangan sistem informasi kesehatan berbasis komunitas (seperti praktik bidan mandiri) dengan pendekatan partisipatif dan teknologi adaptif. Aplikasi yang dihasilkan memiliki potensi untuk digunakan kembali di praktik bidan mandiri lainnya di Indonesia sehingga dapat meningkatkan kualitas pelayanan kesehatan menjadi lebih baik lagi. Kata kunci: rekam medis elektronik; praktik bidan mandiri; sistem informasi kesehatan

  • Comparison of Machine Learning Algorithms for Placement Status Classification : Analysis of Model Accuracy in Supporting Accurate Decision-Making

    by Iin Rohmatika Aulia, Arief Hermawan, Donny Avianto

    Fresh graduates face various challenges in competing for employment opportunities, especially due to the intense competition for limited job positions in the labor market. Both individuals as job seekers and organizations as employers must consider multiple complex factors during the recruitment process. These factors include GPA, university background, aptitude test score, number of projects, internship experience, and academic performance at both the secondary (SSC) and higher secondary (HSC) levels. This study aims to classify job placement eligibility using machine learning methods with algorithms such as Naïve Bayes, Deep Learning, Random Forest, and K-Nearest Neighbors (KNN). The experimental results indicate that the Naïve Bayes algorithm achieved the highest accuracy at 81.90%, followed by Deep Learning with 80.20%. These findings suggest that despite its simplicity, Naïve Bayes is highly effective for datasets with relatively independent and well-distributed features. Meanwhile, other algorithms such as KNN and Decision Tree provided competitive results but showed less consistency in generalizing data. For future research, it is recommended to use a larger and more diverse dataset and to consider additional features such as informal work experience, organizational involvement, or personality data to improve prediction accuracy. Furthermore, exploring ensemble learning or hybrid modeling approaches may lead to more optimal classification outcomes.

  • Open Access

    Analisis Perbandingan Metode DES (Double Exponential Smoothing) dan WMA (Weighted Moving Average) dalam Peramalan Penjualan Laptop

    by Asrul Gunawan, Arief Hermawan, Donny Avianto

    Rapid technological developments increase demand for electronic devices, especially laptops. Fluctuations in monthly sales are a challenge for companies in determining the optimal amount of inventory. The inability to predict market demand can disrupt inventory management and customer satisfaction. Therefore, accurate sales forecasting is essential for planning marketing and procurement strategies. This study compares two sales forecasting methods, namely Double Exponential Smoothing (DES) and Weighted Moving Average (WMA), to analyze the accuracy of each method. The results showed that the DES method has a better level of accuracy with an average MAPE value of 16.72%, compared to WMA which reached 21.22%. This study provides practical insights for companies in choosing the right forecasting method, in order to improve inventory management, product procurement strategies, and customer satisfaction

  • ANALISIS PROPERTI PROSPEK DAN NON-PROSPEK BERDASARKAN DATA IKLAN MENGGUNAKAN METODE K-MEANS DAN K-MEDOIDS

    by Adha Maulana, Donny Avianto

    The significant growth of the property industry market in the city of Yogyakarta. has attracted more attention from property companies. This is glanced at by property companies in bridging sales. However, property agency companies often experience losses in plotting advertising budgets. With this problem, the author offers a clustering system for prospective and non-prospective property spots using the K-Means and K-Medoids methods with the Forward Selection feature selection method. This study aims to allocate advertising budgets to targeted projects with great potential. The data used is primary data, namely IRSC data from company x in Yogyakarta with a range (January 2023–March 2024) totaling 212 records. Data processing uses the RapidMiner application with a data composition of 70% used for training data and 30% for testing data. This process produces a DBI value of 1.060 for the K-Medoids method without feature selection and 1.974 for the K-Medoids method using feature selection. The best method produced by the K-Means method using and without feature selection with a DBI value of 0.148.

  • Open Access

    Identifikasi Penyakit Daun pada Tanaman Solanaceae dan Rosaceae Menggunakan Deep Learning

    by Allan Bil Faqih, Donny Avianto

    Dengan proyeksi populasi global sebesar 9,7 miliar pada tahun 2050, pertanian menghadapi tantangan signifikan dalam memastikan ketahanan pangan. Salah satu kendala utama adalah penyakit tanaman yang menurunkan hasil panen hingga 40% per tahun. Penelitian sebelumnya sering terbatas pada deteksi penyakit pada satu spesies tanaman, sehingga kurang mencerminkan kebutuhan multi-spesies dalam praktik pertanian nyata. Penelitian ini bertujuan untuk mengembangkan dan mengevaluasi sistem deteksi penyakit tanaman berbasis deep learning menggunakan Convolutional Neural Networks (CNN) yang diterapkan pada dua famili tanaman, Solanaceae dan Rosaceae. Dataset yang digunakan adalah PlantVillage, berisi 54.306 citra daun dalam format JPEG yang diunduh dari GitHub, dengan data di luar dua famili tersebut dibuang selama pra-pemrosesan. Tiga model deep learning diuji: transfer learning dengan arsitektur InceptionV3 dan dua CNN kustom (DFE dan LCNN). Model LCNN menunjukkan kinerja terbaik dengan akurasi pelatihan, validasi, dan pengujian masing-masing sebesar 99%, 99%, dan 95%. Sebaliknya, InceptionV3 mencapai akurasi pelatihan 96%, validasi 98%, dan pengujian 92%, sedangkan DFE dengan 86% pelatihan, 94% validasi, dan 82% akurasi pengujian. Analisis confusion matrix menunjukkan kesulitan membedakan antara kentang sehat dan kentang dengan late blight, serta cedar apple rust. Hasil ini menyoroti pentingnya pengembangan arsitektur model spesifik dibandingkan model yang kompleks untuk deteksi penyakit multi-crop secara akurat.


  • Open Access

    Sistem Klasifikasi Kualitas Bunga Cengkeh Kering berbasis Website menggunakan Logika Fuzzy Metode Tsukamoto

    by Arifa Farras Fadhila, Donny Avianto

    Indonesian cloves have strong competitiveness in the main market due to their economic benefits, such as being raw materials for kretek cigarettes, spices, and the perfume industry. However, high global competition demands improvements in product quality and consistency. The manual and subjective sorting of cloves often leads to inaccuracies and inconsistencies in quality, which can be detrimental to farmers, especially in smallholdings. The objective of our research is to develop a web-based system for classifying the quality of dried clove flowers using the Tsukamoto fuzzy logic method. The stages of system development using the waterfall method include system requirements analysis, architecture and interface design, website implementation with the Tsukamoto fuzzy method, and testing. The Tsukamoto fuzzy logic implementation method was chosen due to its ability to process uncertain data and produce consistent output. Our findings successfully produced a web-based system called 'Clove Tester', with an average sensitivity of 45.99% from sensitivity testing based on modifications to the membership function of condition and quality variables. These results indicate that the system has a good adaptability to variations in input data, making it suitable for application to data with a high level of uncertainty or ambiguity in this research.

  • Open Access

    Klasifikasi Penyakit Pada Daun Kopi Robusta Menggunakan Arsitektur AlexNet dan Xception dengan Metode Convolutional Neural Network

    by Nadia Ashari, Donny Avianto

    Diseases on the leaves of robusta coffee plants can have a significant impact on the growth and yield of robusta coffee plants. The leaves of the robusta coffee plant are susceptible to various types of diseases caused by fungi, bacteria or insects with symptoms such as brown, yellow or black patches and discoloration on the surface of the leaves of the robusta coffee plant. Early detection of diseases in robusta coffee leaf plants is very important to obtain effective control to maintain plant health. In this study, a disease classification model on the leaves of robusta coffee plants was made using the Convolutional Neural Network (CNN) architecture. The architecture used in this study is AlexNet and Xception. In this study, a dataset of images of robusta coffee leaves obtained through direct observation of robusta coffee plantations in Temanggung Regency was used. The number of datasets used was 1400 data which was divided into 4 classes, namely healthy, root down, leaf rust and red spider mites. The CNN model was tested by setting parameters consisting of batch size, drop out, learning rate, optimizer and the number of epochs that varied 35, 50 and 100. The results of this study show that the AlexNet architecture model with 50 epoch tests obtains the best accuracy of 98.57% and the Xception architecture obtains an accuracy of 100% in each epoch test. Overall, the use of AlexNet and Xception architectures is very effective in classifying diseases in robusta coffee leaves, but the Xception architecture is superior in the ability to classify complex datasets and higher accuracy.

  • Open Access

    Implementasi Logika Fuzzy Tsukamoto untuk Optimasi Jumlah Produksi Es Batu Kemasan

    by Yurjaa Ghoniyyan Purba, Donny Avianto

    Menentukan jumlah produksi yang akurat merupakan hal penting dalam perencanaan produksi, terutama ketika menghadapi fluktuasi permintaan yang seringkali menjadi tantangan utama. Ketidakpastian dalam permintaan ini memerlukan optimasi agar jumlah produksi yang dihasilkan dapat memenuhi kebutuhan konsumen tanpa menyebabkan penumpukan stok berlebih. Salah satu pendekatan yang dapat digunakan untuk menentukan jumlah produksi adalah metode Logika Fuzzy, khususnya metode Tsukamoto, yang mempertimbangkan variabel permintaan dan persediaan. Penelitian ini bertujuan untuk menerapkan metode Tsukamoto dalam menentukan jumlah produksi es batu kemasan. Data penelitian diperoleh melalui wawancara dengan pemilik usaha dan mencakup data historis produksi, permintaan, serta persediaan es batu kemasan. Hasil penelitian menunjukkan bahwa sistem yang dikembangkan menghasilkan Mean Absolute Percentage Error (MAPE) sebesar 1,86% dan akurasi prediksi sebesar 98,14%. Nilai MAPE yang berada di bawah 10% mengindikasikan bahwa sistem ini mampu memberikan prediksi jumlah produksi yang optimal dan efektif.