Claim Missing Document
Check
Articles

Found 18 Documents
Search

Public Perception of Buying and Selling Bitcoin Using Lexicon Sentiment Analysis Muhammad Rahman Ali; Wijaya, Rifki; Yunanto, Prasti Eko
Indonesia Journal on Computing (Indo-JC) Vol. 9 No. 2 (2024): August, 2024
Publisher : School of Computing, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34818/INDOJC.2024.9.2.980

Abstract

This study investigates public perceptions of Bitcoin (BTC) trading using sentiment lexicon analysis. The rapid growth of cryptocurrency trading has attracted significant public interest and investment, making it crucial to understand the sentiments and opinions surrounding BTC transactions. By employing sentiment lexicon methods, this research analyzes tweets and social media posts to determine public sentiment. The study aims to identify trends and patterns in public opinion, providing insights into how sentiment impacts BTC trading behavior. Preliminary results indicate a correlation between positive sentiment and increased trading activity, while negative sentiment correlates with market declines. This research contributes to a better understanding of the role of public sentiment in the volatile cryptocurrency market.
Multimodal Biometrik pada Keystroke User-Adaptive Feature dan Mahalanobis Distance Hutomo, Ardityo Cahyo Putro Hutomo; Eko Yunanto, Prasti; Sthevanie, Febryanti
Telkatika: Jurnal Telekomunikasi Elektro Komputasi & Informatika Vol. 4 No. 1 (2024): Desember 2024
Publisher : Perpustakaan Universitas Telkom

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

This research analyzes the effectiveness of the combination of User-Adaptive and Mahalanobis Distance methods in biometrics keystroke authentication systems. Using Biomey Keystroke Dataset with 40 respondents, this study aims to improve the accuracy and reliability of KD-based authentication. The developed system consists of enrollment and authentication stages, with User-Adaptive as the feature extraction method and Mahalanobis Distance for feature matching. Decision level fusion technique is applied to integrate the results of various keystroke features. The results obtained show that the fusion technique with Mahalanobis Distance shows better results compared to non-fusion features with an average error reduction of 8.73%. The optimal vector length (Fn) was found at n = 5 with an error value of 12.07%. The best threshold search resulted in a FAR of 15.6% and FRR of 6% at n = 5. The results obtained in this study show a lower error rate with an average decrease in error value of 9.9% with previous studies. This research proves the potential of Mahalanobis Distance and fusion techniques in improving the accuracy of biometrics keystroke authentication systems, opening up opportunities for the development of more reliable security systems. Further studies are recommended to explore certain patterns on the touch screen and the use of more varied datasets and real-time testing data. Keywords- authentication, biometrics
Kmeans-SMOTE Integration for Handling Imbalance Data in Classifying Financial Distress Companies using SVM and Naïve Bayes Maulana, Didit Johar; Siti Saadah; Prasti Eko Yunanto
Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) Vol 8 No 1 (2024): February 2024
Publisher : Ikatan Ahli Informatika Indonesia (IAII)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.29207/resti.v8i1.5140

Abstract

Imbalanced data presents significant challenges in machine learning, leading to biased classification outcomes that favor the majority class. This issue is especially pronounced in the classification of financial distress, where data imbalance is common due to the scarcity of such instances in real-world datasets. This study aims to mitigate data imbalance in financial distress companies using the Kmeans-SMOTE method by combining Kmeans clustering and the synthetic minority oversampling technique (SMOTE). Various classification approaches, including Nave Bayes and support vector machine (SVM), are implemented on a Kaggle financial distress data set to evaluate the effectiveness of Kmeans-SMOTE. Experimental results show that SVM outperforms Nave Bayes with impressive accuracy (99.1%), f1-score (99.1%), area under precision recall (AUPRC) (99.1%), and geometric mean (Gmean) (98.1%). On the basis of these results, Kmeans-SMOTE can balance the data effectively, leading to a quite significant improvement in performance.
Deteksi Varian Penggunaan Helm dari Kamera Surveilans Menggunakan Metode Berbasis Deep Learning Faturahman, Farhan; Yunanto, Prasti Eko; Sulistiyo, Mahmud Dwi
eProceedings of Engineering Vol. 12 No. 3 (2025): Juni 2025
Publisher : eProceedings of Engineering

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Sepeda motor merupakan moda transportasi utama diIndonesia, tetapi tingkat kepatuhan terhadap penggunaanhelm masih rendah. Rekaman kamera surveilans yang seringkali memiliki resolusi rendah menyulitkan deteksi otomatis.Selain itu, variasi helm yang digunakan, seperti full-face,half-face, non-standar, serta pengendara tanpa helm,menjadi tantangan dalam proses pendeteksian. Penelitian inibertujuan untuk mengembangkan model deep learningberbasis YOLOv8n yang mampu mendeteksi penggunaanhelm pada citra beresolusi rendah. Dataset dikumpulkansecara mandiri dengan sudut pandang serta pencahayaanyang serupa. Pengujian dilakukan dengan berbagaikonfigurasi batch size dan jumlah epoch untuk menemukanparameter optimal. Hasil evaluasi menunjukkan bahwamodel dengan 100 epoch dan batch size 32 memberikanperforma terbaik dengan mAP@50 sebesar 0.984, mAP@50-95 sebesar 0.819, precision 0.953, recall 0.952, dan F1-score0.953. Model ini mampu mendeteksi empat kategori helmdengan akurasi tinggi meskipun pada citra beresolusi rendah.Penelitian ini membuktikan bahwa YOLOv8n dapatdigunakan untuk deteksi otomatis penggunaan helm denganakurasi tinggi, yang dapat membantu sistem pemantauan lalulintas dan meningkatkan keselamatan berkendara. Kata kunci: deteksi helm, kamera surveilans, resolusi rendah,deep learning, YOLOv8
Identifikasi Pengguna Berbasiskan Biometrik Keystroke Menggunakan MVMCNN Azzam, Muhammad Abdullah; Yunanto, Prasti Eko; Sulistiyo, Mahmud Dwi
eProceedings of Engineering Vol. 12 No. 3 (2025): Juni 2025
Publisher : eProceedings of Engineering

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Keamanan akses pengguna daring menjadi isukrusial di era digital. Identifikasi berbasis biometrik, sepertikeystroke dynamics, dianggap lebih aman dibandingkanmetode konvensional. Penelitian ini mengimplementasikanMulti-Voter Multi-Commission Nearest Neighbor Classifier(MVMCNN) untuk identifikasi pengguna melalui keystrokedynamics. MVMCNN dipilih karena kemampuannya mengatasikelemahan KNN dengan skema multi-voter dan pendekatanLocal Mean Probabilistic Neural Network (LMPNN). Datasetkeystroke dari Universitas Telkom digunakan dengan fitur UD,DD, DU, UU, dan Duration. Eksperimen meliputi tiga skenario:(1) menentukan panjang vektor optimal (N=4, 8, 12, 16, 20, 24),(2) penyederhanaan fitur menjadi rata-rata dan median, serta(3) seleksi fitur menggunakan Variance Threshold (0.1).Evaluasi menggunakan F1-Score. Hasil menunjukkan skenariopertama dengan N=20 menghasilkan F1-Score tertinggi(0.6911). Penyederhanaan fitur menurunkan performa, denganF1-Score terbaik 0.3031 (mean, k=9) dan 0.3257 (median, k=3),menandakan pentingnya kekayaan informasi dalam fitur.Seleksi fitur menggunakan Variance Threshold tidak banyakmengubah performa, menunjukkan distribusi data sudahoptimal. Temuan ini menegaskan bahwa granularitas databerperan penting dalam akurasi sistem identifikasi berbasiskeystroke dynamics. Kata kunci— biometrik, keystroke, identifikasi, mvmcnn, f1-score.
Sistem Question Answering pada Data Kesehatan Menggunakan Model pre-trained BERT Alhafidz, Bagas Millen; Rachmawati, Ema; Yunanto, Prasti Eko
eProceedings of Engineering Vol. 12 No. 3 (2025): Juni 2025
Publisher : eProceedings of Engineering

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

Setelah pandemi covid-19, kesehatan menjadi halyang harus diperhatikan. Sebagian besar masyarakatmenggunakan search engine sebagai alat untuk mencariinformasi tentang kesehatan. Namun informasi yangdidapatkan berupa query hasil search engine yang masihumum. Sistem Question Answering adalah sistem yangmemberikan informasi sesuai informasi yang dibutuhkan olehpengguna secara spesifik. Pada penelitian ini dibangun sistemQuestion Answering menggunakan metode BidirectionalEncoder Representations from Transformer (BERT). BERTmerupakan sebuah pre-trained model yang menggunakanarsitektur transformer. BERT dapat menyelesaikan tugassistem Question Answering. Dengan pre-trained model, sistemtidak perlu melakukan training model dari awal. Sistem hanyaperlu menggunakan train model yang telah dibuat oleh oranglain sesuai tugas yang dibutuhkan untuk menghemat waktu dansumber daya. Untuk mengukur performansi, digunakan metodeExact Match (EM) dan F1-score. Hasil dari penelitian ini skorterbaik yang didapat yaitu Exact Match 75% dan F1-score76%.Kata kunci— question answering, BERT, pre-trained model,kesehatan
Public Perception of Buying and Selling Bitcoin Using Lexicon Sentiment Analysis Muhammad Rahman Ali; Wijaya, Rifki; Yunanto, Prasti Eko
Indonesian Journal on Computing (Indo-JC) Vol. 9 No. 2 (2024): August, 2024
Publisher : School of Computing, Telkom University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34818/INDOJC.2024.9.2.980

Abstract

This study investigates public perceptions of Bitcoin (BTC) trading using sentiment lexicon analysis. The rapid growth of cryptocurrency trading has attracted significant public interest and investment, making it crucial to understand the sentiments and opinions surrounding BTC transactions. By employing sentiment lexicon methods, this research analyzes tweets and social media posts to determine public sentiment. The study aims to identify trends and patterns in public opinion, providing insights into how sentiment impacts BTC trading behavior. Preliminary results indicate a correlation between positive sentiment and increased trading activity, while negative sentiment correlates with market declines. This research contributes to a better understanding of the role of public sentiment in the volatile cryptocurrency market.
Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) Methods to Forecast Daily Turnover at BM Motor Ngawi Larasati, Larasati; Saadah, Siti; Yunanto, Prasti Eko
Indonesian Journal of Artificial Intelligence and Data Mining Vol 7, No 1 (2024): March 2024
Publisher : Universitas Islam Negeri Sultan Syarif Kasim Riau

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24014/ijaidm.v7i1.27643

Abstract

The number of motorcycles on the report of Indonesian BPS statistics from the Indonesian State Police between 2019 to 2021 by its type has increased annually. Routine motorcycle checks, services, and maintenance are essential to keep a motorcycle in good condition and more durable; therefore, buying spare parts is enlarged in line with the growth of public motorcycle ownership. The necessity of buying spare parts increases with the growth of public motorcycle ownership. Numerous stores in Ngawi offer motorcycle spare parts and check services for routine motorcycle maintenance. One of these stores is BM Motor. To develop an effective product-selling strategy, it is essential to forecast the daily turnover of the shop. To achieve this, the present research aims to analyze the daily turnover using Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM). These methods were applied to a time-series dataset, allowing for an in-depth examination of the patterns and trends in the shop's turnover. The research compares several hyperparameter tunings and scenarios to optimize the models that forecast daily turnover data at the store. The outcomes presented that the LSTM model achieved a lesser MAE score of 0.087, while the RNN model scored 0.092. These findings proved that the LSTM model achieved lower MAE than the RNN model, it means LSTM is more accurate than the RNN model.