Claim Missing Document
Check
Articles

Found 4 Documents
Search
Journal : Jurnal Algoritma

Penggunaan Convolutional Neural Network dan Transfer Learning untuk Rekomendasi Gaya Rambut Pria Azzahra, Monica Salwa; Maesaroh, Syti Sarah; Guntara, Rangga Gelar
Jurnal Algoritma Vol 21 No 2 (2024): Jurnal Algoritma
Publisher : Institut Teknologi Garut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33364/algoritma/v.21-2.2134

Abstract

Hairstyle atau gaya rambut merupakan suatu hal krusial yang mempengaruhi penampilan seseorang. Khususnya pada pria, permasalahan yang sering muncul adalah ketidaksesuaian hasil potongan rambut mereka sehingga menyebabkan rasa percaya diri menurun. Tujuan penelitian ini yaitu mengembangkan sistem rekomendasi gaya rambut bagi pria berdasarkan bentuk wajah dan jenis rambut untuk memberikan rekomendasi yang lebih personal dibandingkan penelitian sebelumnya yang hanya fokus pada satu fitur. Metode yang digunakan pada penelitian ini yaitu Convolutional Neural Network (CNN) dan transfer learning dengan memanfaatkan metode VGG16, fine-tuning, dan extract features dengan fokus pada peningkatan akurasi model. Parameter pelatihan sistem rekomendasi ini menggunakan 32 batch size, 75 epoch, dan learning rate Adam. Model yang digunakan adalah pre-trained model yang didasarkan pada model face shape dan hair type terpilih yang sudah melalui proses pelatihan dan pengujian pada tahap sebelumnya. Hasil pengujian sistem rekomendasi ini menghasilkan dua output, yaitu akurasi face shape model dengan nilai 59,62% dan akurasi hair type model dengan nilai 59,61%. Output tersebut menunjukkan peningkatan nilai akurasi dari penelitian sebelumnya, yang membuktikan bahwa penggunaan metode VGG16 cukup efektif dalam meningkatkan akurasi pada sistem pengolahan gambar terutama pada dataset yang jumlahnya terbatas.
Deteksi Komentar Spam Judi Online Berbahasa Indonesia Menggunakan XGBoost dan TF-IDF Arrayyan, Dzakwan Rafi; Guntara, Rangga Gelar; Nugraha, Muhammad Rizki
Jurnal Algoritma Vol 22 No 2 (2025): Jurnal Algoritma
Publisher : Institut Teknologi Garut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33364/algoritma/v.22-2.3012

Abstract

The phenomenon of online gambling continues to show growth with increasingly worrying trends. One of the challenges faced is the proliferation of gambling promotional comments on the YouTube platform due to the suboptimal performance of spam detection systems in recognizing manipulative language patterns. To address this issue, this study proposes a model for detecting spam comments in Indonesian using a combination of Term Frequency–Inverse Document Frequency (TF-IDF) and Extreme Gradient Boosting (XGBoost). The dataset contains 10,220 YouTube comments that have been manually labeled and processed through preprocessing stages, including unicode normalization and cleaning of irrelevant characters. The model was evaluated using 20% of the test data and produced an accuracy of 91%, precision of 92%, recall of 91%, and an F1-score of 91%. These results show that the combination of TF-IDF and XGBoost is effective for classifying short texts in YouTube comments. Thus, this study contributes to the development of Indonesian-language spam comment detection models, which are still rarely researched, and can also be used as a reference for media platforms in improving the effectiveness of stopping the spread of illegal content through social media comment sections.
Peningkatan Akurasi Rekomendasi Film Menggunakan Neural Collaborative Filtering dengan Arsitektur RecommenderNet Sukmana, Dimas; Guntara, Rangga Gelar; Nugraha, Muhammad Rizki
Jurnal Algoritma Vol 22 No 2 (2025): Jurnal Algoritma
Publisher : Institut Teknologi Garut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33364/algoritma/v.22-2.3013

Abstract

The rapid growth of the film industry and streaming platform users has given rise to the challenge of information overload, where users find it difficult to find films that suit their preferences amid the abundance of content choices. This study aims to develop a Neural Collaborative Filtering (NCF)-based movie recommendation system model with a RecommenderNet architecture to improve prediction accuracy and personal recommendation relevance. The model was evaluated using the Root Mean Square Error (RMSE) metric to assess rating prediction accuracy and Normalized Discounted Cumulative Gain (NDCG@100) to measure recommendation quality and order. The results show that the model achieves an RMSE of 0.1946 and an NDCG@100 of 0.8136, indicating the model's ability to learn user preferences and generate relevant and well-ordered recommendations. This research contributes to the development of more effective and personalized recommendation systems in the digital streaming domain and offers an efficient approach to reducing the impact of information overload and improving the user experience.
Analisis Sentimen Ulasan Pengguna GoPay di Google Play Store menggunakan Model IndoELECTRA Fitriati, Lisna Rahma; Guntara, Rangga Gelar; Purwaamijaya, Btari Mariska
Jurnal Algoritma Vol 22 No 2 (2025): Jurnal Algoritma
Publisher : Institut Teknologi Garut

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33364/algoritma/v.22-2.3053

Abstract

The use of digital wallets in Indonesia continues to increase, making it important to analyze user opinions on services such as GoPay. This study analyzes the sentiment of 9,713 GoPay user reviews from the Google Play Store (September 16–October 2, 2025) using the IndoELECTRA model. The research stages included data collection, pre-processing, tokenization, training, and model evaluation. IndoELECTRA produced 95% accuracy and an F1-score of 0.95, demonstrating its strong performance in processing informal Indonesian text. These results confirm the potential of monolingual models in understanding local linguistic contexts compared to some multilingual models. This research contributes to the development of Indonesian NLP, including sentiment evaluation on neutral ratings, although it is limited to Google Play Store data sources and only covers two main sentiment classes.