Claim Missing Document
Check
Articles

Found 2 Documents
Search

Analisis Komparatif XGBoost dan Temporal Fusion Transformer (TFT) pada Time Series Forecasting Harga Solana Qatrunnada Athirah Herdiyanto; Juhraini Helfiana Lexa; M. Zikry Sahendra Chan
Teknik: Jurnal Ilmu Teknik dan Informatika Vol. 6 No. 1 (2026): Mei : Teknik: Jurnal Ilmu Teknik dan Informatika
Publisher : LPPM Sekolah Tinggi Ilmu Ekonomi - Studi Ekonomi Modern

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51903/teknik.v6i1.1138

Abstract

 Cryptocurrency price prediction, particularly for highly volatile assets like Solana (SOL), is a crucial challenge in time series data analysis in digital finance. This study aims to compare the performance of the XGBoost machine learning algorithm with the Temporal Fusion Transformer (TFT) deep learning model in predicting Solana's daily closing price. The dataset used consists of historical Solana price data and network fundamentals data in the form of Total Value Locked (TVL). The research process includes data preprocessing, dividing training and test data, model training, and evaluation using the Root Mean Squared Error (RMSE) metric. The results show that using the same-day price feature has the potential to cause target leakage, resulting in invalid prediction accuracy. In testing using pure historical data without data leakage, the XGBoost model performed better than TFT with an RMSE of 4.27, while TFT produced an RMSE of 18.59. Furthermore, the integration of network fundamentals data in the form of TVL did not improve prediction accuracy and even caused a decrease in performance for the XGBoost model with an RMSE of 7.10. The results of this study show that the use of historical price action features is more effective than fundamental network indicators for short-term daily Solana price predictions.
Pemodelan Probabilistik Latensi Deteksi Gangguan pada Sistem Monitoring Jaringan Berbasis Polling Menggunakan Simulasi Monte Carlo Dea Anggraini; Zulfahmi Indra; M. Faris Al Rafiq; Qatrunnada Athirah Herdiyanto
Edutik : Jurnal Pendidikan Teknologi Informasi dan Komunikasi Vol. 6 No. 5 (2026): EduTIK : Oktober 2026
Publisher : Jurusan PTIK Universitas Negeri Manado

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.67142/edutik.v6i5.647

Abstract

ABSTRAK  Sistem monitoring berbasis polling memeriksa status perangkat secara berkala, sehingga gangguan yang terjadi di antara dua siklus pengecekan baru terdeteksi pada siklus berikutnya. Penelitian terdahulu umumnya melaporkan waktu respons notifikasi, bukan selang waktu sejak gangguan terjadi sampai peringatan pertama tercatat. Penelitian ini memodelkan selang tersebut, yang disebut latensi deteksi, sebagai variabel acak. Data diperoleh dari 40 percobaan terkendali pada router Mikrotik Cloud Hosted Router yang dipantau sistem berbasis Telegram Bot dengan interval pengecekan 30 detik. Analisis meliputi statistik deskriptif, uji Shapiro-Wilk, pencocokan distribusi Normal dengan maximum likelihood, dan uji Kolmogorov-Smirnov dengan koreksi parametric bootstrap. Simulasi Monte Carlo 10.000 iterasi dijalankan dengan pendekatan parametrik dan bootstrap resampling sebagai validasi silang. Latensi deteksi rata-rata 89,688 detik dengan simpangan baku 8,609 detik. Data tidak menyimpang dari distribusi Normal (Shapiro-Wilk p = 0,4093; Kolmogorov-Smirnov terkoreksi p = 0,5487). Peluang latensi melampaui 90 detik sebesar 0,4832 pada pendekatan parametrik dan 0,4863 pada bootstrap. Pada ambang 100 detik hasil keduanya mulai berbeda (0,1125 dan 0,1560), dan pada 120 detik bootstrap tidak dapat menghasilkan nilai di luar rentang data. Nilai terukur lebih dari dua kali perkiraan awal karena lamanya uji ping saat perangkat tidak merespons. Model parametrik lebih layak untuk ekstrapolasi di luar rentang data teramati. Temuan ini terbatas pada satu interval pengecekan dan lingkungan virtual. ABSTRACT  Polling-based monitoring systems check device status at fixed intervals, so a fault occurring between two cycles is only detected in the next one. Earlier studies mostly report notification response time rather than the interval between a fault and the first recorded alert. This study models that interval, called detection latency, as a random variable. Data came from 40 controlled trials on a MikroTik Cloud Hosted Router monitored by a Telegram Bot-based system with a 30-second check interval. The analysis combined descriptive statistics, the Shapiro-Wilk test, maximum likelihood fitting of a Normal distribution, and a Kolmogorov-Smirnov test with parametric bootstrap correction. A Monte Carlo simulation of 10,000 iterations was run with parametric and bootstrap resampling approaches as cross-validation. Mean detection latency was 89.688 seconds with a standard deviation of 8.609 seconds. The data showed no departure from the Normal distribution (Shapiro-Wilk p = 0.4093; corrected Kolmogorov-Smirnov p = 0.5487). The probability of latency exceeding 90 seconds was 0.4832 under the parametric approach and 0.4863 under bootstrap. At 100 seconds the two diverged (0.1125 and 0.1560), and at 120 seconds bootstrap could not produce values beyond the observed range. The measured value was more than twice the initial estimate because of the ping test duration when the device did not respond. The parametric model is better suited to extrapolation beyond the observed range. These findings are limited to a single check interval and a virtual environment.