Jong Wan Hu

In this study, machine learning approaches were used to predict the electrical conductivity (EC) of conductive cementitious composites under data-limited conditions. Data scarcity was addressed via data augmentation based on a tabular variational autoencoder (TVAE), and multiple machine learning models were trained and evaluated using a consistent set of input variables and validation procedures. Results showed that prediction performance differed across model types. Tree-based and ensemble learning models exhibited stable predictive behavior after data augmentation and hyperparameter optimization, whereas distance-, kernel-, and neural network-based models were more sensitive to changes in data distribution and optimization strategies. TVAE-based data augmentation had a model-dependent effect on prediction. In addition, Shapley additive explanations analysis indicated that the mixing amount of conductive materials and fly ash content were among the most influential factors governing EC, with fly ash content exhibiting nonlinear, value-dependent contribution patterns. These findings provide insights into model-dependent prediction behavior and variable contribution characteristics in EC prediction for conductive cementitious composites under data-limited conditions.

Assessing changes in mean sea level (MSL) has become increasingly critical due to the significance of climate changes. Soft computing techniques are now widely used to reduce the time and cost associated with traditional MSL estimation methods. Historical MSL data is frequently used to predict future values, yet the application of soft computing models to analyze climate change’s impact on MSL remains relatively unexplored. This study aims to develop and compare various soft computing techniques for modeling MSL fluctuations using meteorological data. Random forest (RF), support vector regression (SVR), K-nearest neighbors (KNN) regression, deep neural network (DNN), Gaussian process regression (GPR), and stacked ensemble methods are employed in this study. The newly developed models are statistically assessed for their effectiveness in modeling MSL at Damietta station, Egypt. Variables environmental data such as surface water temperature, pressure, air temperature (average air temperature, dewpoint, wet-bulb, and heat index), and humidity and wind attributes (speed and direction) are utilized and evaluated in modeling MSL. The results indicate that RF, KNN, and GP outperformed other proposed models in modeling MSL during both training and testing phases. The developed weighted stacked ensemble model, integrating RF, KNN, and GPR, outperformed the base models with a correlation coefficient (R) of 0.88 and normalized root mean square error (RMSE) of 0.056 m. MSL modeling at the study station was particularly sensitive to variations in water temperature, wind speed and direction, and atmospheric pressure. This methodology serves as a valuable framework for climate-driven MSL forecasting in developing coastal regions lacking long-term tide records, directly contributing to UNESCO’s Ocean Decade Challenge 5 on coastal resilience.