Domain Knowledge Informed Data Augmentation

Data and codes for the Paper "Leveraging domain knowledge in data augmentation to boost concrete strength prediction accuracy with automated machine learning and deep learning"

1. The data.csv is collected from https://archive.ics.uci.edu/dataset/165/concrete+compressive+strength
2. The data_aug.py provides the hyperparameter optimization process of four data augmentation methods: GaussianCopula, CTGAN, CopulaGAN, and TVAE. The generation of initial and finalized synthetic data is presented in the script.
3. The filter.py defines three anomaly detection methods: RANSAC, IF, and LOF.
4. The automl.py defines the AutoML and AutoDL frameworks.
5. The prediction.py defines the whole experimental process.