Abstract: In addition to Convolutional Neural Network (CNN) applications in the computer vision field, the goal of this presentation is to show how to utilize CNN’s ability to better classify and capture the underlying structures, relationships, and abnormalities on tabular datasets to both complement the performance of machine learning (ML) models and save time plus resources in feature engineering. With big data, CNN’s benefits multiple due to vast different layers, flexible algorithms, and parameters. The presentation is based on a subset of healthcare projects created during the pandemic peak to contribute to underserved communities in Los Angeles County. Data are sourced from Open Data and Public Health of LA County to classify the healthcare services shortage category. The presentation shows the improved results of CNN for multi-class tabular data – as a standalone model, as an ensemble with ML models for diversification and robustness, and as an automatic feature processing framework.
Bio: Ayemya (Mina) Moe is a Marketing Operations Lead at Project Management Institute and also holds Director of Marketing position. She graduated from UCLA with Economics major and is also serving as a reviewer for IEEE CIS (IEEE Transactions on AI).