PhD Defence Weiwei Wang

Keywords: Categorical Data, Data Representation Learning, Knowledge Graph, Graph Embedding
 

"Categorical Data Embedding"

Even the simplest data can sometimes cause problems for data analysis and machine learning models. Take the name of our country, for example. In a dataset, it might appear as "Netherlands", "the Netherlands", or even "Holland".

Weiwei Wang’s research introduces new ways to handle common types of categorical data, like country names or job titles. She presents three creative methods that turn data from tables into graphs, helping to reveal hidden connections between different categories. Special graph techniques are then used to create better representations of these categories.

These new methods were tested on different datasets and consistently performed better than traditional techniques. Weiwei’s work also offers a clear overview of existing methods and practical advice for researchers and professionals. Finally, her research shows how these ideas could be used in real-world fields, such as the pension industry, where it is important to understand complicated categorical data.

Click here for the live stream.