BROWSE
Detail
Conference Paper
A Universal No-Code AutoML Dashboard Builder for Materials Property Prediction
Cited - time in 
Share 
Share
- Authors
- Yo Han Choi, Hyun Jong Kim, Kang Il Choi, Suyoung Chi
- Issue Date
- 2026-08
- Citation
- International Conference on Big Data Applications and Services (BigDAS) 2026, pp.25-27
- Publisher
- 한국빅데이터서비스학회
- Language
- English
- Type
- Conference Paper
- Abstract
- We present a universal, no-code dashboard builder that converts an arbitrary materials experiment or simulation dataset (CSV/Excel) into a deployable property-prediction tool through an eight-step wizard. The system automatically profiles the uploaded data, identifies the material domain, and infers the input (X) and output (Y) factors from multiple structural and semantic signals. It then compares more than eight machine-learning algorithms together with deep neural networks and a tabular foundation model by K-fold cross-validation—reporting R², RMSE, and MAE—and automatically selects the best model with an explainable rationale. The chosen model powers a range-constrained what-if dashboard and is packaged into a fully offline, double-click-runnable bundle. Optional, constraint-preserving data augmentation is also provided. In doing so, it fosters enterprise AI transformation (AX) by turning trained models into reusable assets.
- Keyword
- materials development, AutoML, no-code platform, property prediction, data augmentation, model deployment
- KSP Keywords
- BEST Model, Cross validation(CV), Data Augmentation, Deep neural network(DNN), Learning algorithms, Materials Development, Model deployment, Prediction tool, Property prediction, What-if, k-fold cross-validation