Claim Missing Document
Check
Articles

Found 15 Documents
Search

Implementation of Temporal Fusion Transformer (TFT) for Short-Term Sales Prediction of Telkomsel Data Packages in East Java Muhammad Azkiya Akmal; Trimono; Alfan Rizaldy Pratama
Jurnal Teknologi Informatika dan Komputer Vol. 12 No. 1 (2026): Jurnal Teknologi Informatika dan Komputer
Publisher : Universitas Mohammad Husni Thamrin

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37012/jtik.v12i1.3268

Abstract

The development of the cellular telecommunications industry has driven an increasing demand for fast, stable, and affordable data services. Accurate forecasting of data package sales is a significant challenge for telecommunications operators due to high demand fluctuations and the complexity of time series patterns. This study aims to implement a Temporal Fusion Transformer (TFT) model based on Seasonal-Trend Decomposition using Loess (STL) to predict short-term sales of Telkomsel data packages in East Java. The data used are sales transactions with hourly time resolution from January to June 2024, focusing on the five data packages with the highest transaction volume. The STL method is applied in the pre-processing stage to separate the trend, seasonal, and residual components, which are then used as additional features in the TFT modeling. Model performance is evaluated using Mean Absolute Error (MAE) and Quantile Risk (q-Risk). The results show that the TFT model is able to produce accurate predictions with an MAE value of 3.6941 and an average q-Risk of 0.0808. Furthermore, interpretability analysis revealed that historical sales variables, seasonal components, and calendar variables significantly contributed to the prediction results. These findings indicate that the STL-based TFT approach is effective for short-term sales forecasting and has the potential to support data-driven operational decision-making in the telecommunications sector.
CLASSIFICATION OF HUMAN AND AI-GENERATED INDONESIAN POP SONGS BASED ON SPECTROGRAM USING CONVOLUTIONAL NEURAL NETWORK Faris Nur Tsani; Wahyu Syaifullah Jauharis Saputra; Alfan Rizaldy Pratama
Jurnal INSTEK (Informatika Sains dan Teknologi) Vol 11 No 1 (2026): APRIL
Publisher : Department of Informatics Engineering, Faculty of Science and Technology, Universitas Islam Negeri Alauddin, Makassar, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24252/instek.v11i1.65432

Abstract

Pesatnya perkembangan teknologi kecerdasan buatan (AI) memicu lonjakan produksi lagu generatif yang menyerupai karya manusia, sehingga menghadirkan tantangan signifikan terhadap orisinalitas dan hak cipta musik. Penelitian ini bertujuan mengklasifikasikan lagu Pop Indonesia kategori human-generated dan AI-generated menggunakan pendekatan Convolutional Neural Network (CNN) berbasis arsitektur ResNet-18. Dataset terdiri dari 100 lagu berformat MP3 yang terbagi seimbang antara karya manusia dan karya AI dari platform Suno dan Udio. Data audio diproses melalui teknik segmentasi overlapping window berdurasi 10 detik dengan overlap 5 detik, kemudian diekstraksi menjadi citra spektrogram Short-Time Fourier Transform (STFT). Total data yang dihasilkan mencapai 4.282 segmen audio. Hasil pelatihan selama 100 epoch menunjukkan bahwa model mencapai konvergensi dengan train accuracy 100% dan validation accuracy 95,09%. Pada tahap pengujian menggunakan data yang belum pernah dilihat sebelumnya, model menunjukkan performa unggul dengan tingkat akurasi 93,01%. Temuan ini mengonfirmasi bahwa penggunaan representasi spektogram dalam arsitektur CNN mampu menangkap perbedaan fitur frekuensi dan temporal secara efektif untuk mengidentifikasi musik berbasis AI pada genre Pop Indonesia.
EfficientNetB4–Vision Transformer Fusion for Chili Leaf Disease Classification Using Multi-Source Datasets Angga, Reza Putri; Saputra, Wahyu Syaifullah Jauharis; Pratama, Alfan Rizaldy
bit-Tech Vol. 8 No. 3 (2026): bit-Tech
Publisher : Komunitas Dosen Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32877/bt.v8i3.3753

Abstract

Chili plants are a commodity susceptible to plant pest organism attacks that can significantly reduce productivity. Visual identification of chili diseases by farmers is often inaccurate due to symptom similarity across disease categories, necessitating a technology-based approach capable of performing classification automatically and accurately. This study proposes a hybrid model combining EfficientNetB4 and Vision Transformer for chili leaf disease classification into four categories healthy, yellowish, curl leaf, and spot leaf. EfficientNetB4 extracts local features through compound scaling and MBConv blocks, while ViT models global relationships among image regions through self-attention, enabling a semantically meaningful integration of local and global feature representations that addresses the individual limitations of CNN and transformer-based architectures. The dataset integrates 4,000 secondary images from GitHub and 800 primary images collected directly from chili cultivation fields in Central Java, with splitting performed separately per source to ensure proportional distribution across subsets. To evaluate generalization capability, the model was assessed across three scenarios: training and testing on secondary data only 98.25%, testing on primary field data without prior field exposure 87.50%, and training and testing on integrated data 99.17%, with a perfect accuracy of 100% on the primary-only test set. These results demonstrate that incorporating field-collected data into training directly bridges the generalization gap caused by domain shift between laboratory and real-world conditions, outperforming both single-architecture and previous hybrid approaches reported in prior studies. The findings provide a methodological foundation for developing robust automated disease detection systems applicable across diverse agricultural crops and real-world farming environments.
VERIFIKASI OTOMATIS SERTIFIKAT MENGGUNAKAN IMPROVED SIFT DAN RANSAC BERBASIS ANALISIS VISUAL LOGO DAN STEMPEL Maulidya Prastita Syah; Wahyu Syaifullah Jauharis Saputra; Alfan Rizaldy Pratama
Jurnal INSTEK (Informatika Sains dan Teknologi) Vol 11 No 1 (2026): APRIL
Publisher : Department of Informatics Engineering, Faculty of Science and Technology, Universitas Islam Negeri Alauddin, Makassar, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24252/instek.v11i1.66776

Abstract

Proses verifikasi sertifikat prestasi pada jalur Seleksi Nasional Berdasarkan Prestasi (SNBP) masih dilakukan secara manual dan menghadapi tantangan akibat tingginya jumlah dokumen serta variasi kualitas citra. Kondisi ini meningkatkan risiko kesalahan dalam menilai kredibilitas sertifikat, khususnya pada elemen visual seperti logo dan stempel yang tidak valid. Penelitian ini bertujuan mengembangkan metode otomatis dalam mengidentifikasi kredibilitas sertifikat berbasis analisis fitur visual. Metode yang digunakan adalah Improved Scale Invariant Feature Transform (Improved SIFT) untuk mengekstraksi fitur lokal yang stabil, serta Random Sample Consensus (RANSAC) untuk menyaring kecocokan fitur berdasarkan konsistensi geometris. Pencocokan dilakukan antara citra sertifikat dan citra parameter berupa logo dan stempel yang telah diidentifikasi sebagai tidak valid. Pengujian dilakukan pada 200 citra sertifikat yang terdiri dari dua kelas, yaitu kredibel dan tidak kredibel yang menunjukkan Improved SIFT mampu meningkatkan kualitas representasi fitur dengan akurasi sebesar 0.47 dibandingkan SIFT standar sebesar 0.16. Selanjutnya, RANSAC mampu menyaring false match dengan peningkatan akurasi dari 0.47 menjadi 0.91 setelah RANSAC. Kontribusi penelitian ini adalah mengintegrasikan Improved SIFT dan RANSAC untuk meningkatkan validitas pencocokan fitur visual dalam mendeteksi indikasi ketidak-kredibelan sertifikat pada citra dengan variasi kualitas tinggi. Dengan demikian, metode yang diusulkan mampu meningkatkan keandalan sistem dalam proses verifikasi otomatis berbasis analisis visual.
Sharpe Ratio-Based Dynamic Crypto Asset Allocation with Trend Filtering Using SMA Fauzan Adziima, Andri; Wara, Shindi Shella May; Nasrudin, Muhammad; Pratama, Alfan Rizaldy
Enthusiastic : International Journal of Applied Statistics and Data Science Volume 6 Issue 1, April 2026
Publisher : Universitas Islam Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.20885/enthusiastic.vol6.iss1.art1

Abstract

This paper proposes a dynamic cryptocurrency asset allocation strategy that combines Sharpe Ratio-based weighting with trend filtering using the Simple Moving Average (SMA) of Bitcoin (BTC). The model reallocates capital among a portfolio of seven major cryptocurrencies (BTC, ETH, BNB, SOL, TON, TRX, XRP) every three days, conditional on BTC trading above its respective SMA threshold (50-day, 100-day, or 200-day). When BTC trends below the SMA, the strategy shifts fully to USDT to minimize downside risk. Using historical data from January 1, 2024, to January 1, 2025, the study evaluates performance across three SMA configurations and benchmarks against a buy-and-hold baseline. Results show that the SMA-50 strategy achieved the highest cumulative return (+231.51%) and Sharpe Ratio (2.51), significantly outperforming both the longer SMA-based models and the baseline average return (+132.14%). Risk analysis indicates that shorter SMA windows allow more responsive exposure during market uptrends but increase short-term volatility. Overall, the findings support the use of hybrid strategies combining trend-following filters and risk-adjusted allocation for managing crypto portfolios in volatile environments.