Surya Agustian
Universitas Islam Negeri Sultan Syarif Kasim Riau, Pekanbaru

Published : 5 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 5 Documents
Search

Penerapan Support Vector Machine dan FastText untuk Mendeteksi Hate Speech dan Abusive pada Twitter Afdhal Zikri; Surya Agustian
JURNAL MEDIA INFORMATIKA BUDIDARMA Vol 7, No 1 (2023): Januari 2023
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/mib.v7i1.5408

Abstract

Hate speech and abusive language are negative tendencies that often appear on social media recently. In addition, due to the advancement of technology and the rapid growth of the internet, anyone can now engage in hate speech or even offensive expression such as in Twitter, which eventually leads to fights on that social media platforms. Automatic detection of offensive contents and hate speech is recommended to be applied, especially on the user application’s side, to filter tweet contents which destruct social life in the real world. The purpose of this research is to create a classification model using Support Vector Machine with FastText word embeddings features, to classify if a tweet contains hate speech and/or offensive language. Our contribution in this research is an improvement in performance from the baseline SVM (support vector machine) with FastText word embeddings features input. The experiment results will also be compared with several machine learning method that have been reported using the same dataset of 13,167 tweets. The experiment using the most optimal SVM model, yields an average accuracy of 82.65%, with the accuracies of the hate speech class, abusive language class and hate speech level, are 84.92%, 86.60% and 76.43% respectively. These results are better than conventional machine learning, but cannot exceed the results achieved by deep learning.
Klasifikasi Citra Stroke Menggunakan Augmentasi dan Convolutional Neural Network EfficientNet-B0 Nadila Handayani Putri; Jasril Jasril; Muhammad Irsyad; Surya Agustian; Febi Yanto
JURNAL MEDIA INFORMATIKA BUDIDARMA Vol 7, No 2 (2023): April 2023
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/mib.v7i2.5981

Abstract

A stroke is a sudden onset of brain dysfunction, lasting for 24 hours or longer, resulting from clinically focal and global brain dysfunction. As many as 15 million people die from stroke each year. The stroke patients need an immediate treatment to minimize the risk of brain damage. One of the proponents for the stroke diagnosis is through a computed tomography (CT) image. In recent years, the image processing techniques capable to detect stroke patterns in a brain image, it can be useful for doctors and radiologists in doing diagnosis and treatment. This study aims to compare the level of accuracy using augmentation and without augmentation and hyperparameters using the Convolutional Neural Network in the EfficientNet-B0 architecture to classify ischemic, hemorrhagic, and normal brain stroke images. The data augmentation is produced by rotating, horizontal flipping, and contrast tuning of the original data. Testing data is provided as much as 20% of the portion of the original and augmented data, and the other 80% is used for the training process to find the optimal model. The model search is based on the composition of the training and validation data with a ratio of 70:30, 80:20 and 90:10. The experimental results show that the best performance is obtained for the combined original and augmented images, with accuracies of 97%, 93%, and 94%, respectively, for the three types of data-test: original, augmented, and combined. The merging of original and augmentated images for training data has shown that the model is robust enough in producing high accuracy results.
Perbandingan Klasifikasi Citra CT-Scan Kanker Paru-Paru Menggunakan Contrast Stretching Pada CNN dengan EfficientNet-B0 Alfitra Salam; Febi Yanto; Surya Agustian; Siti Ramadhani
KLIK: Kajian Ilmiah Informatika dan Komputer Vol. 4 No. 3 (2023): Desember 2023
Publisher : STMIK Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/klik.v4i3.1448

Abstract

Data from the World Health Organization (WHO) indicates that in 2020, approximately 10 million people died from cancer. Smoking has been identified as a primary factor causing lung cancer, as cigarettes contain over 60 toxic substances that can trigger the development of the disease. The rate of lung cancer has rapidly increased due to excessive cigarette consumption. Detecting nodules in the lungs typically takes about 10-30 minutes. In this study, a Convolutional Neural Network (CNN) algorithm with EfficientNet-B0 architecture is employed to classify lung cancer. The preprocessing process involves contrast stretching, and various hyperparameter optimization techniques such as Adam, Adagrad, and SGD are used to enhance the CNN's performance. Average pooling with output dense layers of 64, 32, 16, 1 is utilized. Performance analysis is conducted using a confusion matrix. The highest classification results are achieved using the ADAM optimizer with a learning rate of 0.01, where accuracy reaches 72.48%, precision is 71.52%, recall is 64.2%, and the F1 score is 64.76%. Meanwhile, results obtained from the original dataset show differences. The highest classification result is obtained using the ADAM optimizer with a learning rate of 0.01, achieving an accuracy of 64.22%, precision of 52.69%, recall of 50.52%, and an F1 score of 43.51%. These results indicate that the use of contrast stretching in lung cancer classification preprocessing is highly effective in improving accuracy
Perbandingan Klasifikasi Citra CT-Scan Kanker Paru-Paru Menggunakan Image Enhancement CLAHE Pada EfficientNet-B0 Dzaky Abdillah Salafy; Febi Yanto; Surya Agustian; Fitri Insani
KLIK: Kajian Ilmiah Informatika dan Komputer Vol. 4 No. 3 (2023): Desember 2023
Publisher : STMIK Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/klik.v4i3.1514

Abstract

In recent years, there has been a significant increase in the global cancer related mortality rate. Among various cancer types, lung cancer has emerged as one of the highest incidence cases. Lung cancer predominantly affects males and is attributed to several factors, including exposure to cigarette smoke, long-term air pollution, and exposure to carcinogenic compounds such as radon, asbestos, arsenic, coal tar, and diesel fuel emissions. The growth of cancerous cells in the lungs can be detected using various imaging techniques, with CT-Scan being one of them. This research focuses on the classification of normal lung organs and those affected by cancerous cells. The classification process employs two types of data: original data and data processed with Contrast Limited Adaptive Histogram Equalization (CLAHE). The data is initially divided with 90:10 ratios before being trained using a Convolutional Neural Network (CNN). The CNN architecture used is EfficientNet-B0, with the assistance of different optimizers and learning rates. After testing, the model's performance is evaluated using a confusion matrix to compare the results between the use of original data and CLAHE-processed data. The use of CLAHE processed data yields higher evaluation metrics compared to the original data, achieving a precision of 87.9%, recall of 85.6%, F1-score of 85.11%, and accuracy of 85.29% in the 90:10 data split, with the Adam optimizer and a learning rate of 10-1. The research results reveal that the utilization of image enhancement, specifically Contrast Limited Adaptive Histogram Equalization (CLAHE), with an appropriate combination of clip limit and tile grid, can impact the model's performance in classifying image data.
Klasifikasi Sentimen Masyarakat di Twitter Terhadap Ganjar Pranowo dengan Metode Support Vector Machine Syaiful Azhar; Yusra; Muhammad Fikry; Surya Agustian; Iis Afrianty
KLIK: Kajian Ilmiah Informatika dan Komputer Vol. 4 No. 3 (2023): Desember 2023
Publisher : STMIK Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/klik.v4i3.1537

Abstract

The classification of public sentiment towards Ganjar Pranowo on Twitter can provide insights into his popularity, support, or criticism. This research aims to classify public sentiment towards Ganjar Pranowo on Twitter using the Support Vector Machine (SVM) method. The research data consists of 4000 tweets collected from Twitter. After undergoing preprocessing, these tweets are classified using SVM into positive or negative classes. The classification method is optimized to produce the most optimal model by testing the influence of feature selection stages and SVM parameter tuning. The data is divided into 80% training (TRAIN_SET) and 20% testing (TEST_SET). The optimal model is validated using 10% of the randomly selected TRAIN_SET for validation data. Sixteen experiments are conducted to explore the optimal model, with the highest validation results (top rank 4 models) tested on the TEST_SET, yielding F1-scores of 84.13%, 84.13%, 84.13%, and 84.13% for experiment IDs 1, 7, 14, and 16, respectively. In this research, SVM proves to be sufficiently effective in classifying sentiment-related tweets about Ganjar Pranowo on Twitter