Announcing the Candidacy of Randyll Pandohie for the PhD in Big Data Analytics
This study advances the field of one-class classification by developing a novel methodology for sequential multiple change point detection in time series data. Traditional batch multiple change point methods often require the entire dataset to be available and are typically processed in a single batch. In contrast, our approach emphasizes the sequential detection of multiple change points, allowing for real-time processing and immediate anomaly detection in evolving data streams. We employ the Least Squares-Support Vector Regression (LS-SVR) coupled with the LS-SVDD algorithms due to their proficiency in handling non-linear patterns and robustness in anomaly identification. This methodology is rigorously evaluated through a series of experiments on simulated time series datasets, designed to mimic a wide range of change point scenarios and a real world dataset. Our results showcase the capability of our approach to accurately detect multiple change points sequentially, presenting a significant improvement over the traditional batch processing techniques and opening new avenues for real-time data monitoring across various domains.
Committee in Charge:
Dr. Edgard Maboudou (Chair)
Dr. Alexander Mantzaris
Dr. Hsin-Hsiung Huang
Dr. Tingting Zhang (Rosen College of Hospitality Management)
Read More