Knight Foundation School of Computing and Information Sciences
Leila Zahedi is a Ph.D. candidate in the Sustainability, Optimization, and Learning for InterDependent networks (solid) lab at the Knight Foundation School of Computing and Information Sciences. She holds a B.S. in Computer Engineering and a M.S. in Information Technology Management: Advanced Information Systems from the University of Isfahan and Yazd University, Iran, respectively. Her research interests include optimization algorithms for automated machine learning, data science, and educational data mining. She is currently working under Dr. M. Hadi Amini’s supervision, focusing on automated machine learning optimization. She was awarded the Student Government Association (SGA) Graduate Student Scholarship, Grace Hopper, and Richard Tapia scholarships, among others. She published more than 10 journal and conference papers. Besides her academic activities, she also served as the chair of the Graduate and Professional Student Committee (GPSC) and graduate senator at SGA.
Machine learning (ML) is an evolving branch of computational algorithms that allows computers to learn from experience and historical datasets. It is one of the fast-growing research fields and has garnered much attention from academic and industrial researchers to discover the patterns and data representations from the raw data. It addresses manipulating, managing, mining, and understanding different data types to solve real-world challenges in different fields such as banking, transformation, and education. However, building a useful machine learning model is a challenging process that requires consideration of both time and performance.
Overall, machine learning’s main objective involves determining the best and most predictive model, which can be obtained by suitable feature engineering and hyper-parameters tuning. This is where the novel idea of automated machine learning (AutoML) emerged to automate the process of applying ML techniques and find optimal machine learning solutions and release the burden of huge development costs and manual processes. The challenges regarding ML models’ complicated and computationally expensive nature create significant opportunities for future research avenues on these steps, specifically on Hyper-Parameter Optimization (HPO) and Feature Selection Optimization (FSO) methods which are among the primary tasks in AutoML. The proposed approach will be implementing an automated system that addresses these issues and provides complete automatic steps to make well-performed time-efficient prediction models.