Software Engineering Department Discusses Doctoral Thesis on Interpretable Intelligent Prediction of Hospital Services
Duhaa Fadill Abbas
As part of its ongoing scientific activities, the Software Engineering Department at the College of Information Technology – University of Babylon, hosted on Sunday, September 21, 2025, the defense of the doctoral thesis by student Mohammad Khudair Abbas, entitled "Interpretable Dynamic Prediction of Hospital Services Using an Intelligent Model", under the supervision of Prof. Dr. Iman Saleh Sakban, in the college’s conference hall.
The thesis focused on developing an interpretable dynamic deep learning model (DIDL) for predicting the quality of healthcare services in hospitals, leveraging a bidirectional long short-term memory (Bi-LSTM) architecture. The model aims to address the challenges of prediction in healthcare, where accuracy and reliability are crucial to ensuring the consistent provision of high-quality services.
The study involved comprehensive data preprocessing, including cleaning, deduplication, feature aggregation, selection, transformation, handling missing values, and applying collaborative filtering and Kalman filter techniques to enhance prediction accuracy. Two types of datasets were utilized: the first collected from Iraqi hospitals across Babel, Karbala, Baghdad, and Diwaniya, and the second obtained from U.S. hospitals via the Centers for Medicare & Medicaid Services (CMS).
Results demonstrated the model’s superior performance, achieving an accuracy of 98.4% with the U.S. datasets and 95.8% with the Iraqi datasets, outperforming traditional models such as LSTM, Bi-LSTM, and regression techniques. This reflects the model’s effectiveness in enhancing predictions while minimizing error rates.
The study underscores the significance of integrating interpretable deep learning approaches in healthcare service prediction, contributing to improved transparency, reliability, and decision-making in hospital operations, ultimately supporting optimal patient outcomes.