GANs on Synthetic Tabular Data Generation

Navigating class imbalances in classification problems is a recurring challenge. often impeding the accurate classification of minority classes within the data.  While there are many oversampling and undersampling techniques available, they often falter in grasping the intricacies of the minorities, relying primarily on local information to create synthetic samples.  The sharing aims to demonstrate the untapped potential of Generative Adversarial Networks (GANs) in the realm of synthetic tabular data generation, which remains as an underexplored frontier as compared to the image domain.

Key Takeaways

• Delve into the enduring challenge of class imbalances in classification problems and the difficulties that machine learning models face in accurately classifying minority classes
• Showcase the transformative potential of GANs as a solution to address class imbalances in classification problems as compared to existing oversampling/undersampling techniques
• Highlight the impact of GANs in model performance and discuss potential valuable gains that can be achieved

—————————————————————————————————————————

Bio

Grace Ngu Sook Ern | Senior data scientist | Micron Semiconductor Asia Pte Ltd | Singapore

With 5 years of extensive experience spanning analytics, machine learning, and deep learning within oil and gas and manufacturing sectors, Grace has spearheaded and meticulously nurtured a series of transformative machine learning projects. Each of these culminated into successful deployment into production, boasting millions of savings and reduction in greenhouse gas emissions. Grace’s experience in project productionalization encompasses a wide array of machine learning and deep learning techniques, including classification, regression, anomaly detection, clustering, computer vision, model monitoring and statistical techniques, applied to multifaceted business domains. Beyond project success, she is committed to staying updated with the latest developments in data science, and currently has a strong interest in the field of Generative AI.