Claim Missing Document
Check
Articles

Deep Learning-Based Lung Sound Classification Using Mel-Spectrogram Features for Early Detection of Respiratory Diseases Yabani, Midfai; Faisal, Mohammad Reza; Indriani, Fatma; Nugrahadi, Dodon Turianto; Kartini, Dwi; Satou, Kenji
Journal of Electronics, Electromedical Engineering, and Medical Informatics Vol 8 No 1 (2026): January
Publisher : Department of Electromedical Engineering, POLTEKKES KEMENKES SURABAYA

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35882/jeeemi.v8i1.1256

Abstract

Respiratory diseases such as asthma, chronic obstructive pulmonary disease, and pneumonia remain among the leading causes of death globally. Traditional diagnostic approaches, including auscultation, rely heavily on the subjective expertise of medical practitioners and the quality of the instruments used. Recent advancements in artificial intelligence offer promising alternatives for automated lung sound analysis. However, audio is an unstructured data format that must be converted into a suitable format for AI algorithms. Another significant challenge lies in the imbalanced class distribution within available datasets, which can adversely affect classification performance and model reliability. This study applied several comprehensive preprocessing techniques, including random undersampling to address data imbalance, resampling audio at 4000 Hz for standardization, and standardizing audio duration to 2.7 seconds for consistency. Feature extraction was then performed using the Mel Spectrogram method, converting audio signals into image representations to serve as input for classification algorithms based on deep learning architectures. To determine optimal performance characteristics, various Convolutional Neural Network (CNN) architectures were systematically evaluated, including LeNet-5, AlexNet, VGG-16, VGG-19, ResNet-50, and ResNet-152. VGG-16 achieved the highest classification accuracy of the tested models at 75.5%, demonstrating superior performance in respiratory sound classification tasks. This study demonstrates the potential of AI-based lung sound classification systems as a complementary diagnostic tool for healthcare professionals and the general public in supporting early identification of respiratory abnormalities and diseases. The findings suggest that automated lung sound analysis could enhance diagnostic accessibility and provide more valuable support for clinical decision-making in respiratory healthcare applications
Gender Classification of Twitter Users Using Convolutional Neural Network Fitra Ahya Mubarok; Mohammad Reza Faisal; Dwi Kartini; Dodon Turianto Nugrahadi; Triando Hamonangan Saragih
MATRIK : Jurnal Manajemen, Teknik Informatika dan Rekayasa Komputer Vol. 23 No. 1 (2023)
Publisher : Universitas Bumigora

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30812/matrik.v23i1.3318

Abstract

Social media has become a place for social media analysts to obtain data to gain deeper insights and understanding of user behavior, trends, public opinion, and patterns associated with social media usage. Twitter is one of the most popular social media platforms where users can share messages or ”tweets” in a short text format. However, on Twitter, user information such as gender is not shown, but without realizing it or not, there is information about it in an unstructured manner. In social media analytics, gender is one of the important data that someone likes, so this research was conducted to determine the best accuracy for gender classification. The purpose of this study was to determine whether using combined data can improve the accuracy of gender classification using data from Twitter, tweets, and descriptions. The method used was word vector representation using word2vec and the application of a 2D Convolutional Neural Network (CNN) model. Word2vec was used to generate word vector representations that take into account the context and meaning of words in the text. The 2D CNN model extracted features from the word vector representation and performed gender classification. The research aimed to compare tweet data, descriptions, and a combination of tweets and descriptions to find the most accurate. The result of this study was that combined data between tweets and
Performance Analysis of the Fuzzing Method in Detecting API Vulnerabilities in Mobile Healthcare Application X Based on OWASP API Security Top 10 Muhammad Ikhwanul Hakim; Radityo Adi Nugroho; Dodon Turianto Nugrahadi; Rudy Herteno; Setyo Wahyu Saputro
Telematika Vol 19, No 1: February (2026)
Publisher : Universitas Amikom Purwokerto

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.35671/telematika.v19i1.3149

Abstract

Traditional perimeter security measures, such as Web Application Firewalls (WAFs) and static analysis, often fail to detect logic-based vulnerabilities in healthcare Application Programming Interfaces (APIs), creating significant risks for patient data confidentiality. Addressing the scarcity of empirical performance evaluations in this domain, this study employs a grey-box controlled experimental design to assess the effectiveness of automated HTTP fuzzing against a production-grade mobile health application ("Application X"). Using the FFUF tool configured with sequential identifier injection, status-code filtering, and hidden-field probing, the experiment tested 33 endpoints against the OWASP API Security Top 10 2023 benchmarks. To ensure data reliability, a rigorous multi-step validation protocol including replay testing and environmental noise elimination was applied to filter false positives. The results identified 88 distinct vulnerabilities distributed across six categories, with a critical dominance of Security Misconfiguration (API8) and Broken Object Property Level Authorization (API3). Analytically, the high prevalence of API3 reveals a systemic failure in backend serialization, where sensitive fields  including password hashes and internal administrative flags were exposed due to the absence of Data Transfer Objects (DTOs), contradicting the assumption of secure client-side filtering. Limitations of this study include the restriction to a single patient-role perspective and the exclusion of third-party integrations. The study concludes that automated fuzzing is superior to static analysis in detecting runtime data leakage and recommends mandatory Server-Side Output Filtering through explicit DTOs as a critical standard for secure health API development and data privacy compliance.
Dampak dari Parameter Variasi Koneksi, Node dan Kecepatan Node Terhadap Delay pada Routing Protocol AODV dan BATMAN Jaringan MANET Dodon Turianto Nugrahadi; M Reza Faisal; Liling Triyasmono; Muhammad Janawi
Jurnal Komputasi Vol. 8 No. 2 (2020)
Publisher : Jurusan Ilmu Komputer Fakultas MIPA Universitas Lampung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.23960/komputasi.v8i2.2675

Abstract

Mobile ad-hoc Network (MANET) is a multihop wireless network that a many collection of mobile nodes that are dynamic. MANET each node on the network have the same position, so it needs the appropriate routing protocol, to support the exchange of data to be optimal. In this study, the routing protocol to be tested is AODV and BATMAN based scenario increasing the number of connections, nodes and speed. Simulation parameter scenarios is number connection 1 UDP, 2 UDP, 3 UDP, and number of node 25 node, 50 node, 100 node, and then number node speed 20 m/s, 50 m/s. in this AODV routing protocol will establish a rute from the source node to the destination only if there is a request from the source node. BATMAN routing protocols, all decisions and information disseminated throughout the node and will regularly update on each node. The performance parameters to be measured such as delay by using OMNET ++ 4.6. Output of simulation will analysis with two way anova and multivariate to know correlation between variation scenario impact to delay. The results obtained in this study AODV and BATMAN have their respective advantages, analisys with two-way anova show that both AODV and BATMAN get the impact of the scenario from incrising the number of connections, the number of nodes and the number of nodes speed with a p-value of 0.012212 (<0.05) with two-way anova. From all scenarios, the number of UDP connections has the greatest impact, from UDP 1, UDP 2 and UDP 3. Followed by the number of speed 50 and node 100. So it can be concluded that the connection has an effect on increasing delay. The increasing number of speed and nodes can contribute to an increase in delay if number of nodes above 100 and speed above 50. With multivariate analysis, the BATMAN protocol had the most impact on the delay under the scenario then AODV.
Analisis Komparasi Implementasi Steganografi White-Space dan White-Space Modified pada Artikel Terenkripsi AES dalam HTML5 Rudy Herteno; Dodon Turianto Nugrahadi; Muhammad Sholih Afif; M Reza Faisal; Friska Abadi
Jurnal Komputasi Vol. 8 No. 1 (2020)
Publisher : Jurusan Ilmu Komputer Fakultas MIPA Universitas Lampung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.23960/komputasi.v8i1.2525

Abstract

The level of internet usage continues to increase until now.  information exchange requires security that cannot be predicted by others.  one technique for securing information is steganography.  Steganography techniques are the science and art of hiding information.  This technique can hide the content of information in media that cannot be guessed by ordinary people, so as not to arouse suspicion of the people who see it.  One of the media that can implement the white-space modified steganography method is HTML pages.  in addition, AES (Advanced Encryption Standard) is a lighter encryption security algorithm compared to other algorithms. In this study, plain text that has been encrypted into cipher text is then inserted with white-space and white-space modification steganography techniques. Data changes have occurred but only less than 1 percent.  In experiments that have been implemented on Google Chrome and Mozilla Firefox are the same except in Internet Explorer, which changes the data slightly larger.The implementation of AES encryption and stegano white-space original, has 100% success but the 80% decryption process is successful, but the decryption results contain additional binaries. This happen because the use of tabulation (tabs) instead of spaces in HTML5 articles, and this is often found in HTML articles. while the implementation of AES encryption and stegano whitespace modified, has a success of 100% and the decryption process of 90% succeeded without any changes. 1 article failed because the number of articles is too small compared to the amount of space provided. The conclusion that implementation of AES encryption and white-space modified is more appropriate to be implemented in HTML5 articles, and than the use of tabulation and the number of characters also consequences on the implementation.Keywords: Information, Steganography, White-space modified, Security, AES, Web Browser 
Studi Ekstraksi Fitur Berbasis Vektor Word2Vec pada Pembentukan Fitur Berdimensi Rendah irwan budiman; M Reza Faisal; Dodon Turianto Nugrahadi
Jurnal Komputasi Vol. 8 No. 1 (2020)
Publisher : Jurusan Ilmu Komputer Fakultas MIPA Universitas Lampung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.23960/komputasi.v8i1.2517

Abstract

Klasifikasi teks adalah salah satu metode untuk mengelola dan mencari informasi penting yang terdapat pada format tekstual yang tidak terstruktur. Ekstraksi fitur merupakan proses penting pada klasifikasi teks untuk mengubah format tekstual yang tidak terstruktur menjadi terstruktur sehingga dapat diproses oleh algoritma machine learning untuk mengklasifikasikan ke class yang telah ditentukan. Salah satu teknik ekstraksi fitur yang umum digunakan adalah vector space representation. Teknik ini mudah digunakan tetapi berpotensi menghasilkan data dengan dimensi banyak yang berakibat kepada peningkatan waktu komputasi bahkan tidak dapat diproses karena limitasi perangkat keras. Pada riset ini kami melakukan studi terhadap teknik ekstraksi fitur yang mampu menghasilkan data berdimensi sedikit. Ekstraksi fitur yang digunakan memanfaatkan vektor word2vec untuk mengontrol jumlah fitur yang dihasilkan. Pada riset ini kami membandingkan beberapa model yang dihasilkan sendiri dengan jumlah fitur yang bervariasi dan model yang telah disedikan oleh Google. Hal ini dilakukan untuk mengetahui jumlah fitur yang dapat menghasilkan kinerja klasifikasi terbaik. Hasilnya didapat nilai kinerja tertinggi akurasi yaitu 0.877 dengan jumlah fitur adalah 300 dari model yang dihasilkan sendiri.
Perbandingan Nilai K pada Klasifikasi Pneumonia Anak Balita Menggunakan K-Nearest Neighbor Dwi Kartini; Andi Farmadi; Muliadi muliadi; Dodon Turianto Nugrahadi; Pirjatullah Pirjatullah
Jurnal Komputasi Vol. 10 No. 1 (2022)
Publisher : Jurusan Ilmu Komputer Fakultas MIPA Universitas Lampung

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.23960/komputasi.v10i1.2965

Abstract

Pneumonia adalah penyakit menular yang menyerang saluran pernapasan bagian bawah dan merupakan salah satu penyebab utama kematian pada anak-anak di bawah lima tahun. Pneumonia mudah menyerang balita yang disebabkan oleh berbagai mikroorganisme yang ada di lingkungan seperti virus, bakteri, jamur dan bakteri mikro. Penelitian ini menggunakan K-Nearest Neighbor (KNN) untuk klasifikasi pneumonia pada pasien berdasarkan gejala yang dialami. Metode klasifikasi KNN dilakukan dengan membandingkan jarak objek antara data tes dan objek keseluruhan pada data pelatihan berdasarkan data riwayat medis pasien. Perbandingan persentase data pelatihan dan data pengujian yang digunakan adalah 90:10, 80:20, dan 70:30 untuk menghitung nilai jarak terdekat dari data pengujian dengan data pelatihan keseluruhan dengan jumlah k yang digunakan. Matriks kebingungan digunakan untuk mengukur hasil tes klasifikasi Pneumonia untuk balita dengan kombinasi jumlah data pelatihan dan data pengujian pada jumlah nilai k = {1, 3, 5, 7, 9, 11}, akurasi tertinggi, presisi, penarikan, dan nilai ukuran-F diperoleh. 0,86, 0,89, 1, dan 0,91 untuk data pelatihan 90%, 10% data pengujian dengan nilai k = 3.
Co-Authors Abadi, Friska Abdul Gafur Adi Mu'Ammar, Rifqi Adi, Puput Dani Prasetyo Adi, Puput Dani Prasetyo Ahmad Rusadi Ahmad Rusadi Ahmad Rusadi Arrahimi - Universitas Lambung Mangkurat) Ahmad Rusadi Arrahimi - Universitas Lambung Mangkurat) Aida, Nor Aji Triwerdaya Alfando, Muhammad Alvin Andi Farmadi Andi Farmadi Andi Farmadi Andi Farmadi Ando Hamonangan Saragih Apriana, Susi Ardiansyah Sukma Wijaya Arfan Eko Fahrudin Arifin Hidayat Azwari, Ayu Riana Sari Azwari, Ayu RianaSari Bachtiar, Adam Mukharil Badali, Rahmat Amin Bahriddin Abapihi Bedy Purnama Cahyadi, Rinova Firman Dike Bayu Magfira, Dike Bayu Djordi Hadibaya Dwi Kartini Dwi Kartini Dwi Kartini, Dwi Emy Iryanie, Emy Faisal Murtadho Faisal, Mohammad Reza Fajrin Azwary Fatma Indriani Fhadilla Muhammad Fitra Ahya Mubarok Fitria Agustina fitria Fitriani, Karlina Elreine Fitrinadi Friska Abadi Gunawan Gunawan Gunawan Gunawan Halim, Kevin Yudhaprawira Hariyady, Hariyady Herteno, Rudy Herteno, Rudy Heru Kartika Candra, Heru Kartika Huynh, Phuoc-Hai Ichsan Ridwan Indah Ayu Septriyaningrum Irwan Budiman Irwan Budiman Irwan Budiman Irwan Budiman Ismail Didit Samudro Julius Tunggono Jumadi Mabe Parenreng Junaidi, Ridha Fahmi Kartika, Najla Putri Keswani, Ryan Rhiveldi Kevin Yudhaprawira Halim Liling Triyasmono M Kevin Warendra M. Apriannur Martalisa, Asri Maulidha, Khusnul Rahmi Mera Kartika Delimayanti Miftahul Muhaemen Muhamad Ihsanul Qamil Muhammad Alkaff Muhammad Anshari Muhammad Haekal Muhammad Hasan Muhammad Ikhwanul Hakim Muhammad Irfan Saputra Muhammad Itqan Masdadi Muhammad Itqan Mazdadi Muhammad Janawi Muhammad Khairin Nahwan Muhammad Mirza Hafiz Yudianto Muhammad Nazar Gunawan Muhammad Reza Faisal, Muhammad Reza Muhammad Rofiq Muhammad Sholih Afif Muhammad Solih Afif Muliadi Muliadi Muliadi MULIADI -, MULIADI Muliadi Aziz Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi Muliadi, M Musyaffa, Muhammad Hafizh Nafis Satul Khasanah Nahdhatuzzahra Nahdhatuzzahra Ngo, Luu Duc Noor Hidayah Nursyifa Azizah Ori Minarto Padhilah, Muhammad Pirjatullah Pirjatullah Pirjatullah Prastya, Septyan Eka Priyatama, Muhammad Abdhi Radityo Adi Nugroho Rahayu, Fenny Winda Rahmad Ubaidillah Rahmat Ramadhani, Rahmat Ramadhan, Muhammad Rizky Aulia Riadi, Putri Agustina Rifki Izdihar Oktvian Abas Pullah Rifki Riza Susanto Banner Rizal, Muhammad Nur Rizki Amelia Rizki, M. Alfi Rozaq, Hasri Akbar Awal Rudy Herteno Rudy Herteno Rudy Herteno Saman Abdurrahman Saputro, Setyo Wahyu Saputro, Setyo Wahyu Saputro, Setyo Wahyu Saragih, Triando Hamonangan Satou, Kenji Selvia Indah Liany Abdie Setyo Wahyu Saputro sholih Afif Siti Napi'ah Soesanto, Oni Sri Cahyo Wahyono Sri Rahayu Sri Redjeki Sri Redjeki Totok Wianto Totok Wiyanto Tri Mulyani Triando Hamonangan Saragih Umar Ali Ahmad Utomo, Edy Setyo Wahyu Dwi Styadi Wahyu Saputro, Setyo Wardana, Muhammad Difha Winda Agustina Yabani, Midfai Yanche Kurniawan Mangalik YILDIZ, Oktay Yudha Sulistiyo Wibowo Zamzam, Yra Fatria