A Ph.D. Awarded to Researcher Hiba Mohammed Ahmed from the Department of Computer Science at the College of Computer
Researcher Hiba Mohammed Ahmed Al-Marwa’i earned her Ph.D. in Computing and Information Science, specializing in Computer Science, from the Department of Computer Science at the College of Computer and Information Technology, Sana’a University. Her dissertation, entitled “A Novel Multi-Objective Metaheuristic Model for Selecting Optimal Set of Features from Bioinformatic Big Data” was successfully defended on Sunday, 17 Sha’ban 1446 AH (corresponding to February 16, 2025).
The examination and evaluation committee consisted of:
•Professor Dr. Ghalib Hamoud Al-Ja’fari – Principal Supervisor and Committee Member
•Associate Professor Mousa Masleh Ghurab – Internal Examiner
•Professor Dr. Munir Abdullah Saeed Hazaa – External Examiner and Chair of the Committee
The dissertation aimed to develop new models for identifying the most significant genetic variations associated with Alzheimer’s disease from DNA sequences. Given that most high-dimensional gene expression data contain a vast number of redundant genes, this poses challenges to machine learning algorithms due to the high dimensionality. Feature selection has proven to be an effective method for enhancing the performance of classification algorithms by addressing two core goals: reducing the number of features and improving classification accuracy.
The primary objective of this research was to propose an efficient hybrid multi-objective method for feature selection. The method integrates the TOPSIS technique—a multi-attribute decision-making approach—along with a filtering mechanism to extract informative features, and the multi-objective Crow Search Algorithm (CSA) to simultaneously minimize the number of features and classification error. Through the use of Opposition-Based Learning (OBL), the study mitigated the risk of CSA converging on locally optimal solutions. To evaluate the model’s effectiveness, the researcher conducted experiments on standard microarray datasets from the ADNI database, comparing the results with six single-objective and three multi-objective methods. The findings show that the proposed approach outperforms single-objective methods in classification accuracy, and it demonstrates superior performance over other multi-objective algorithms in terms of classification accuracy and the number of selected features.
Among the recommendations presented in the dissertation is implementing the proposed model in discrete space and comparing those results with those obtained in continuous space.
A number of academics, researchers, students, interested attendees, as well as the researcher’s colleagues and family members, were present at the defense.
