Ziyi Liu

Driven and accomplished educator from Monash University, adept in Python Programming and fostering student mastery in complex subjects. Demonstrates exceptional organizational skills and a commitment to continuous improvement. Proven track record in enhancing learning experiences through innovative teaching strategies, with a passion for Machine Learning and Statistical Analysis.

• Concept Drift Detection and Handling in Time Series (Supervised by Christoph Bergmier)

The purpose of this project is to investigate the feasibility and model performance of using different concept detectors to detect concept drift in time series and handle these concept drift. Because forecasting always uses regression methods, the classification methods should be relatively easily adaptable to regression, and the ensemble method is a common way to deal with non-stationary data classification. However, there are some drawbacks to these methods. Therefore, this project aims to find an ensemble method to detect and handle concept drift.

• Smoothed Piece-wise Constant Function:\\ A Probabilistic Kernel Smoothing Model (Supervised by Daniel Schmidt and Mario Boley)

This paper introduces a novel approach employing a kernel smoothing method applied probabilistically to non-smooth piece-wise constant functions, diverging from traditional methods that typically smooth observed variables. By focusing on smoothing the prediction function itself, this method not only enhances the adaptability and smoothness of a model but also provides the analytical solution of uncertainty quantification, thereby boosting overall model reliability in complex environments.