BIOS : Jurnal Teknologi Informasi dan Rekayasa Komputer

Klasifikasi Sentimen Pengguna Platform X Terhadap Isu Geopolitik Israel–Palestina Menggunakan BERT

2026-10-09T19:35:59+07:00

The geopolitical issue between Israel and Palestine has sparked a wide range of public reactions on social media, particularly on the platform X. Sentiment analysis is a widely discussed topic in Natural Language Processing. This study develops a sentiment classification model for that issue using the BERT Base Uncased architecture, with data labeled automatically through the VADER approach. Data were collected via scraping with Tweet-Harvest, yielding 3.172 tweets, later reduced to 2.362 after preprocessing stages including cleaning, case folding, tokenizing, stopword removal, and stemming. VADER labeling produced a distribution of 73.7% negative sentiment and 26.3% positive sentiment. The model achieved 86.92% accuracy, with precision, recall, and F1-Score for the negative class of 90%, 93%, and 91%, respectively, while the positive class obtained 77%, 71%, and 74%. The average F1-score reached 88%. These results indicate the model correctly predicts most of the data, although a slight indication of overfitting appeared, with no significant impact on validation accuracy. This confirms that combining VADER and BERT Base Uncased is effective for sentiment classification on international geopolitical issues.

Prediksi Kurs Rupiah terhadap Dolar dan Yuan Menggunakan LSTM-PSO dan XGBoost

2026-10-14T00:02:02+07:00

The Indonesian rupiah exchange rate is a key determinant of Indonesia's economic conditions and is vulnerable to global geopolitical dynamics, yet geopolitical risk is still rarely incorporated explicitly into exchange rate forecasting models. This study compares the accuracy of XGBoost and Long Short-Term Memory optimized with Particle Swarm Optimization (LSTM-PSO) in forecasting IDR/USD and IDR/CNY, using the Geopolitical Risk Daily (GPRD) index as an external predictor. The dataset comprises daily IDR/USD, IDR/CNY, and GPRD observations from February 2022 to February 2026. The workflow includes data collection, cleaning, and integration, feature engineering, normalization, chronological data splitting, and model evaluation using RMSE, MAE, MAPE, and R². Under the tested scenario, LSTM-PSO yielded lower errors than XGBoost for both currency pairs, with a MAPE below 1% and an R² above 0.89, whereas XGBoost produced a negative R². These findings indicate the potential of LSTM-PSO as an exchange rate forecasting model that incorporates geopolitical risk information, and may serve as a consideration for banks, investors, and international businesses in monitoring exchange rate movements.

RC-MH-GRU untuk Prediksi Arah Pandang 360° pada Telepresensi Robot

2026-10-16T10:28:21+07:00

Viewport prediction can help a 360-degree video system prioritize panoramic regions for tile-based, region-of-interest (ROI), or adaptive-quality streaming. This study evaluates a Ridge-calibrated multi-horizon gated recurrent unit (RC-MH-GRU) offline at 100, 200, 300, and 500 ms. The aggregate dataset contains 72 videos and 2,986 head-tracking sessions sampled at 10 Hz. Each two-degree-of-freedom viewing direction is mapped to a three-dimensional unit vector; one second of position and first-difference history forms the input. Ridge-AR and the 14,604-parameter MH-GRU are combined on the unit sphere with one validation-selected weight per horizon, yielding 15,336 parameters. Evaluation covers five training seeds, a video-disjoint test split, five-fold video cross-validation, strong baselines, ablation, computational profiling, video-level statistics, and a parameter-matched 14,412-parameter Tiny Transformer. Across 60,000 test windows, mean angular errors are 0.958°, 2.033°, 3.324°, and 6.041°, lower than Ridge-AR at all horizons and compared with 1.026°, 2.090°, 3.383°, and 6.118° for the Tiny Transformer. Cross-validation on 72 videos also favors RC-MH-GRU over Ridge-AR at 200–500 ms. The evidence supports a compact prediction component for prospective tile/ROI allocation, not bandwidth savings, viewport-quality gains, or end-to-end telepresence performance.