Comparison of Classification Models for Predicting Admission Outcomes of Prospective Students with Disabilities

Rosihon Anwar, Mohamad Irfan, Ilham Nurjaman — Tue, 31 Mar 2026 00:00:00 +0000

Students with disabilities are a group that requires special attention in the admission process at universities, especially at State Islamic Higher Education Institutions (PTKIN). Although inclusive policies have been implemented, challenges in implementation in the field are still quite significant, especially in terms of equal access and the readiness of educational institutions. This study aims to analyze the opportunities and challenges of accepting students with disabilities at PTKIN through a machine learning approach to predict the factors that influence selection graduation. The research data consists of 80 prospective students with disabilities who participated in the PTKIN selection, covering variables such as gender, province of origin, previous education, school accreditation, and type of disability. The research process included data cleaning, feature engineering (including categorical encoding and recategorization of disability variables), and data balancing using the SMOTE method. Next, model training was carried out using three main algorithms, namely Support Vector Machine (SVM), Random Forest, and XGBoost, as well as model combination (ensemble voting classifier) for performance comparison. The results show that the SVM (RBF kernel) model provides the best performance with an accuracy of 80% and an F1-score of 0.88 for the “Pass” class. This model outperforms Random Forest and XGBoost, which have an accuracy of 65% each. The most influential factors for graduation are the province of origin, disability category, and previous form of education. These findings indicate that the acceptance of students with disabilities at PTKIN is still influenced by geographical factors and educational background, so affirmative policies need to be directed at expanding access for people with disabilities from certain regions and backgrounds. The machine learning approach has proven to be effective as a tool for analyzing inclusive education policies in the PTKIN environment.

Comparison of Long Short-Term Memory and Recurrent Neural Network For Stock Market Price Movement Classification in Islamic Bank Finance

Mon, 13 Apr 2026 00:00:00 +0000

This study addresses the importance of accurate stock price prediction in the Islamic finance sector, where reliable forecasting supports better investment decisions and market stability. Despite the growing use of deep learning methods, comparative studies on sequential models in this domain remain limited. Therefore, this research compares the performance of Long Short-Term Memory (LSTM) and Recurrent Neural Network (RNN) models for classifying stock price movement direction of Islamic banks in Indonesia. The dataset was sourced from two Islamic banks in Indonesia, covering the period from 2022 to mid-2024, with features such as Open, High, Low, Close, Adjusted Close, and Volume. The CRISP-DM method was applied for data processing, and testing was performed with data splits of 60:40, 70:30, and 80:20, as well as epoch variations (30, 50, 80). Results indicate that RNN outperforms LSTM, with the highest accuracy of 58% for RNN and 53% for LSTM. Evaluation metrics also included precision, recall, and F1-score. In conclusion, RNN performs better for stock movement classification direction, while LSTM is more effective for minimizing prediction error.

CoreID Journal

Comparison of Classification Models for Predicting Admission Outcomes of Prospective Students with Disabilities

Comparison of Long Short-Term Memory and Recurrent Neural Network For Stock Market Price Movement Classification in Islamic Bank Finance