Claim Missing Document
Check
Articles

Found 2 Documents
Search
Journal : Jurnal Pseudocode

Klasifikasi Level Non-Proliferatif Retinopati Diabetik Dengan Ensemble Convolutional Neural Network Ruvita Faurina; Endina Putri Purwandari; Mario Tiara Pratama; Indra Agustian
Jurnal Pseudocode Vol 8, No 1 (2021): Volume 8 Nomor 1 Februari 2021
Publisher : Universitas Bengkulu

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (438.898 KB) | DOI: 10.33369/pseudocode.8.1.1-10

Abstract

Penelitian ini mengusulkan algoritma CNN ensemble classifier untuk klasifikasi level non-proliferatif Retinopati diabetik. Penelitian ini menggunakan metode transfer learning feature-extraction, dan membandingkannya dengan fine-tuning. Pada lapisan pertama lapisan klasifikasi, dibandingkan penggunaan lapisan GAP dan Flatten dengan menggunakan metode dropout. Mode terbaik digunakan sebagai mode final klasifikasi. Arsitektur yang digunakan adalah DenseNet201, InceptionV3 dan MobileNetV2, Masing-masing model diuji dengan optimasi SGD dan ADAM. Keputusan prediksi diambil berdasarkan metode average voting. Hasil pengujian masing-masing arsitektur menunjukkan hasil terbaik adalah fine tuning, GAP, dan optimasi ADAM. Model final fine-tuning DenseNet201, InceptionV3 dan MobileNetV2 dapat mengklasfikasi level retinopati diabetik dengan akurasi pada data uji masing-masing 93%, 94% dan 89%. Sedangkan performa klasifikasi model ensemble untuk masing-masing kelas memiliki akurasi terendah 95,6% dan F1-Score terendah 91.3%.Kata Kunci: retinopati diabetik, deep learning, convolutional neural network, ensemble classifier, DenseNet201,  InceptionV3, MobileNetV2.
Sentiment Analysis Komentar Berbahasa Bengkulu Menggunakan Long Short-Term Memory (LSTM) Safitri Nurhaeni; Ruvita Faurina; Ferzha Putra Utama; Kurnia Anggriani
Jurnal Pseudocode Vol 10 No 2 (2023): Volume 10 Nomor 2 September 2023
Publisher : UNIB Press

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.33369/pseudocode.10.2.117-125

Abstract

Sentiment information is one type of information that can be obtained from social media. Sentiments can be interpreted as opinions and views of society which contain feelings. To analyze the value of sentiment whether the sentiment is a sentiment that tends to be neutral, negative, or positive, sentiment analysis can be used. language has its characteristics and uniqueness, Bengkulu language is no exception, because of this, it is necessary to model sentiment analysis for various languages. Sentiment modeling for the Bengkulu language is not yet available, therefore a sentiment analysis model for the Bengkulu language is developed by applying Long Short-Term Memory (LSTM), and architectural experiments for Long Short-Term Memory (LSTM) are carried out to obtain an architectural sentiment analysis model that produces the best value. The data used in the study amounted to 24,000 Bengkulu-language comments received from social media Instagram, Twitter, and Youtube. Experimental research 1 produces the best accuracy value compared to the results of testing in other experiments, with an accuracy value of 0.87 a precision value of 0.80, a recall value of 0.82, and an F1 score of 0.81 Keywords: Information, sentiment, Long Short- Term Memory (LSTM), Bengkulu Language, architecture, social media.