Claim Missing Document
Check
Articles

Found 29 Documents
Search

Perbandingan Kinerja Model Pembelajaran Mesin Random Forest dan K-Nearest Neighbor (KNN) untuk Prediksi Risiko Kredit pada Layanan Pinjaman Online Prayudani, Santi; Sibarani, Yous; Salam, Azrizal; Lubis, Arif Ridho
Journal Software, Hardware and Information Technology Vol 5 No 2 (2025)
Publisher : Jurusan Sistem Informasi Universitas Islam Negeri Alauddin Makassar

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24252/shift.v5i2.204

Abstract

This study aims to compare the performance of two popular machine learning algorithms, Random Forest and K-Nearest Neighbor (KNN), in predicting creditworthiness in online lending systems. The research uses the publicly available Loan Approval Prediction Dataset from Kaggle, which contains borrower profiles such as employment status, number of dependents, annual income, loan amount, loan term, and credit score. Data preprocessing included cleaning, handling missing values, outlier removal, and transformation through normalization and encoding. The dataset was divided into 80% training data and 20% testing data. Random Forest was configured with 100 decision trees and unlimited depth, while KNN used an optimal k value of 5 determined by grid search. Model performance was evaluated using accuracy, precision, recall, and F1-score. The results showed that Random Forest outperformed KNN with consistently higher values (97%) across all metrics, demonstrating strong stability and superior pattern recognition capabilities. KNN, with an accuracy of 89%, still showed good performance and can be considered a lightweight alternative for simpler applications.
Optimization of Deep Learning Algorithms for Medical Image Detection in Cloud Computing-Based Health Applications Putri, Desfita Eka; Prayudani, Santi; Sitopu, Joni Wilson
Journal of Artificial Intelligence and Development Vol. 4 No. 1 (2025): Journal of Artificial Intelligence and Development
Publisher : Edujavare Publishing

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

Abstract

The integration of deep learning into cloud-based healthcare systems has opened new frontiers in medical image analysis, enabling faster, more accurate, and accessible diagnostics. However, the high computational demands of conventional deep learning models pose significant challenges for deployment in cloud environments, especially in latency-sensitive and resource-limited settings. This study aims to optimize deep learning algorithms to enhance their efficiency and scalability for medical image detection within cloud computing infrastructures. A quantitative research approach was employed, involving algorithmic optimization techniques such as pruning, quantization, transfer learning, and federated learning. The models were tested using benchmark medical image datasets and deployed in a simulated cloud environment to evaluate performance metrics such as accuracy, inference time, resource usage, and privacy compliance. Results showed that optimized models, particularly EfficientNet with pruning and quantization, achieved high diagnostic accuracy (up to 91.7%) while significantly reducing computational overhead. Federated learning proved effective in maintaining data privacy with minimal loss in accuracy. The findings suggest that lightweight, secure, and fast deep learning models can be realistically integrated into cloud-based healthcare applications. This study contributes a framework for efficient and scalable AI deployment in clinical settings, particularly in underserved or remote areas.
Prediction of Cyberbullying in Social Media on Twitter Using Logistic Regression Prayudani, Santi; Adha, Lilis Tiara; Ariyani, Tika; Lubis, Arif Ridho
Journal of Applied Informatics and Computing Vol. 9 No. 4 (2025): August 2025
Publisher : Politeknik Negeri Batam

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30871/jaic.v9i4.9842

Abstract

As cases of cyberbullying on social media increase, there is a need for efficient measures to detect the vice. This research aims to establish the application of machine learning algorithms in analyzing text on social media to determine potentially harmful comments using logistic regression. The first and most important research question of this study is to assess the extent to which the model is capable of correctly identifying the comments that contain features of cyberbullying and those that do not. The data set included comments from different social media sites and was preprocessed before further analysis was conducted on it. Exploratory Data Analysis was applied in the study to establish relationships and textual features with bullying behavior. As with any other model, after training and testing the model, the results were analyzed using parameters like precision, precision, gain, and F1 statistics. The outcomes of this study revealed that the use of logistic regression models can give a fairly satisfactory level of accuracy in identifying cyberbullying. In light of this, this study underscores the need to use machine learning algorithms to minimize negative actions in cyberspace.
Peningkatan Kapasitas Serapan Pakan Hijauan Guna Mereduksi Biaya Pengadaan Pakan Kambing Di Desa Tanjung Gusta Aminuddin, Harris; Suadi; Hidayat, Ahmad; Prayudani, Santi
Jurnal Pengabdian Kepada Masyarakat dan Desa Volume 2, Nomor 2, Januari 2025
Publisher : Politeknik Negeri Medan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51510/passa.v2i2.2645

Abstract

Mitra memiliki 56 ekor kambing yang terdiri dari 35 kambing domba (gibas) dan 21 kambing kacang. Dua jenis kambing ini diberi pakan rumput yang didapat dari lahan kosong milik warga yang tidak produktif. Dalam sehari mitra harus mendapatkan 6 ikat rumput yang beratnya ±30 kg/ikat. Biaya pengadaan pakan rumput Rp 15.000/ikat atau Rp 90.000/hari. Selain rumput diberikan pakan tambahan berupa konsentrat secara terpisah untuk menambah daya tahan kambing terhadap penyakit. Mitra menyediakan konsentrat sebanyak 50 kg yang harganya Rp 180.000 untuk 10 hari. Jika dihitung, biaya untuk konsentrat sebesar Rp 18.000/hari, sehingga biaya pembelian pakan sebesar Rp Rp 108.000/hari. Biaya sebesar itu cukup berat, sementara rumput yang diberikan tidak semua dimakan disebabkan sistem pemberian pakan yang tidak tepat. Pemberian rumput tanpa dicacah menyebabkan banyak rumput jatuh ke tanah, terinjak-injak dan bercampur dengan kotorannya yang jumlahnya mencapai 50%. Jika dikonversikan, mitra kehilangan Rp 45.000/hari atau Rp 1.350.000/bulan. Oleh sebab itu, untuk mengurangi atau bahkan meniadakan kerugian tersebut dengan mencacah rumput yang akan diberikan ke kambing. Mesin pencacah rumput adalah solusi pilihan, karena adanya mesin semua rumput yang dberikan termakan tanpa sisa. Dengan mengkonsumsi rumput cacah secara maksimal kambing akan cepat tumbuh kembang gemuk, sehingga masa tunggu layak jual tidak terlalu lama.
Tree Triple Exponential Smoothing Analysis in Forecasting of Fertilizer Sales Prayudani, Santi; Banjarnahor, Wiwin Sry Adinda; Nugroho, Muhammad Rivan; Tazkiyatun Nisa
Acceleration, Quantum, Information Technology and Algorithm Journal Vol. 1 No. 2 (2024): VOLUME 1, NO 2: DECEMBER 2024
Publisher : Yayasan Asmin Intelektual Berkah

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.62123/aqila.v1i2.49

Abstract

The majority of Indonesia's population relies on the agricultural sector, making fertilizer an essential raw material to increase productivity. PT. Pupuk Iskandar Muda (PIM), faces challenges in maintaining the balance of urea fertilizer production and demand. In 2021, PIM's urea fertilizer production was unable to meet demand, while in 2019, 2020, and 2022 there was overproduction. This inventory non-optimization can lead to productivity bottlenecks and increased storage costs. One solution to this problem is forecasting. This research uses the Triple Exponential Smoothing (TES) forecasting method in forecasting urea fertilizer sales for the next period. The data used is fertilizer sales data from PT PIM for the 2019-2023 period. Evaluation of the accuracy value is done using the MAD, MSE, and MAPE matrices. The results of this study indicate that the TES method with a smoothing weight value of Alpha = 0.4, Beta = 0.2, and Gamma = 0.4 produces a MAD value of 22,017.75, MSE of 990,752,983.08, and MAPE of 22.3% which can be categorized as quite feasible to use in forecasting the demand for urea fertilizer at PT PIM seen from the MAPE value.
Multimodal Sentiment Analysis in Indonesian: A Comparative Study of Deep Learning Models for Hate Speech Detection on Social Media Muhammadiyah, Mas’ud; Xiang, Yang; Na, Li; Nishida, Daiki; Prayudani, Santi
Journal International of Lingua and Technology Vol. 4 No. 1 (2025)
Publisher : Sekolah Tinggi Agama Islam Al-Hikmah Pariangan Batusangkar, West Sumatra, Indonesia.

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.55849/jiltech.v4i1.824

Abstract

With the rapid expansion of social media, the prevalence of hate speech has become a critical issue, particularly in the context of Indonesian language and culture. The detection of hate speech in social media platforms is a complex task due to the multimodal nature of online communication, where text, images, and videos are often combined to express sentiments. This study aims to explore and compare deep learning models for multimodal sentiment analysis, focusing on their effectiveness in detecting hate speech in Indonesian social media content. By analyzing both textual and visual data, the study seeks to enhance the accuracy of sentiment classification, specifically identifying instances of hate speech. The research employs several state-of-the-art deep learning models, including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Transformer-based models, to perform sentiment analysis on a multimodal dataset. The dataset includes text and images from Indonesian social media posts, labeled for hate speech detection. The results show that multimodal models outperform text-only models, with the Transformer-based model yielding the highest accuracy and F1-score in detecting hate speech. The inclusion of visual data significantly improved the model’s ability to classify complex and subtle expressions of hate speech. This study concludes that multimodal deep learning models offer a promising solution for detecting hate speech in Indonesian social media, with implications for better content moderation and online safety.
Pengaruh Penerapan Metode Student Centered Learning (SCL) Dan Discovered Learning (DL) Mengenai Pemahaman Mahasiswa Pada Pembelajaran Teknologi Informasi Prayudani, Santi
MULTINETICS Vol. 5 No. 1 (2019): MULTINETICS Mei (2019)
Publisher : POLITEKNIK NEGERI JAKARTA

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32722/multinetics.v5i1.1464

Abstract

Model pembelajaran saat ini sangat bervariasi, mahasiswa dapat mengeksplorasi berbagai ide maupun wawasan yang dimiliki dan mampu mengembangkan wawasan berpikir secara bebas tanpa pernah merasa terpenjarkan pola pikirnya dengan menggunakan metode pembelajaran Student Centered Learning (SCL). Selain itu, mahasiswa juga diharuskan untuk mengeksplorasi dan mengidentifikasi masalah yang muncul sehingga mereka dapat menemukan pengetahuan sendiri. Pembelajaran ini dikenal sebagai Discovery Learning (DL) dimana model pembelajaran ini merupakan model penyelesaian masalah yang akan sangat berguna bagi siswa dalam menghadapi kehidupan nyata di masa depan. Penelitian ini bertujuan untuk mengetahui pengaruh penerapan metode Discovery Learning (DL) pada pemahaman siswa tentang pembelajaran teknologi informasi. Hasil studi dari kedua model pembelajaran tersebut diharapkan dapat meningkatkan kualitas pendidikan di perguruan tinggi khususnya di bidang teknologi informasi dan dapat mengembangkan bidang keilmuan Model pembelajaran saat ini sangat bervariatif, mahasiswa dapat mengeksplorasikan ide maupun gagasan yang dimiliki serta mampu mengembangkan wawasan berpikir secara bebas tanpa pernah merasa terpenjarakan pola pikirnya dengan menggunakan metode pembelajaran Student Centered Learning (SCL) selain itu mahasiswa juga dituntut untuk dapat menggali serta mengidentifikasikan permasalahan yang muncul sehingga mereka dapat menemukan pengetahuan dengan sendirinya. Pembelajaran ini dikenal dengan pembelajaran penemuan atau Discovery Learning (DL) dimana model pembelajaran ini merupakan suatu model pemecahan masalah yang akan sangat bermanfaat bagi mahasiswa dalam menghadapi kehidupan nyata di kemudian hari. Penelitian ini bertujuan untuk mengetahui pengaruh penerapan metode Discovery Learning (DL) mengenai pemahaman mahasiswa pada pembelajaran teknologi informasi. Hasil penelitian dari kedua model pembelajaran tersebut diharapkan dapat meningkatkan mutu pendidikan di perguruan tinggi khususnya di bidang teknologi informasi serta dapat mengembangkan bidang keilmuan tersebut.
The Application of Artificial Intelligence in Processing Health Data in Biomedical Information Prayudani, Santi; Lase, Yuyun Yusnida; Husna, Meryatul; Adam, Hikmah Adwin
Journal of Computer Science Advancements Vol. 3 No. 2 (2025)
Publisher : Yayasan Adra Karima Hubbi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.70177/jsca.v3i2.2245

Abstract

The increasing complexity and volume of health data in modern biomedical systems have necessitated advanced technologies for effective data processing and analysis. Traditional methods often fall short in managing real-time, multidimensional data generated from various biomedical sources, such as electronic health records (EHRs), wearable devices, and genomic data. This research investigates the application of artificial intelligence (AI) in optimizing the processing and interpretation of biomedical health data. The objective of this study is to explore how AI-based technologies, including machine learning and deep learning algorithms, enhance the efficiency, accuracy, and predictive capabilities in biomedical information systems. By identifying patterns, anomalies, and correlations in large datasets, AI offers potential improvements in disease diagnosis, patient monitoring, and treatment personalization. This research employs a qualitative systematic review method, analyzing peer-reviewed literature published between 2015 and 2024 from major databases such as PubMed, IEEE Xplore, and Scopus. The analysis focuses on case studies, comparative evaluations, and implementation outcomes of AI in various biomedical domains. The findings reveal that AI applications significantly improve data processing speed and accuracy, enable early diagnosis of diseases such as cancer and diabetes, and support predictive analytics for patient outcomes. However, challenges remain in areas such as data privacy, ethical compliance, and algorithm transparency. In conclusion, the integration of AI into biomedical data systems holds transformative potential for healthcare delivery, though further interdisciplinary collaboration is required to address its limitations and ensure equitable access and ethical use.
Analysis of Regression and Neural Network Models in Predicting Patient Visit Volume Harizahahyu; Friendly; Fathoni, Muhammad; Lase, Yuyun Yusnida; Prayudani, Santi; Harfita, Nur Laily
International Journal of Science and Society Vol 7 No 4 (2025): International Journal of Science and Society (IJSOC)
Publisher : GoAcademica Research & Publishing

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.54783/ijsoc.v7i4.1561

Abstract

Predicting patient visit volume plays a crucial role in supporting decision-making and resource allocation in healthcare services. This study aims to compare the performance of Multiple Linear Regression and an Artificial Neural Network (ANN) in forecasting patient visits at a dental clinic, using daily patient visit data and predictor variables such as holidays and promotional activities. Multiple regression was used to capture the linear relationship between the predictor and response variables, while ANN was applied to explore potential non-linear relationships. The results indicate that multiple regression outperformed the ANN, demonstrated by lower Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values, and provided clearer interpretability, making it more beneficial for healthcare practitioners, particularly in the context of a limited dataset. In contrast, the ANN tended to produce overestimates and was less responsive to short-term variations. Therefore, multiple regression can still be considered a reliable, efficient, and interpretable prediction method for clinical data with a moderate sample size, while future research is recommended to use larger datasets and test other machine learning algorithms to improve the accuracy and generalizability of the results.