Automated Data Engineering and Feature Selection for the Case Study of Warpage Detection in Fused Deposition Modeling

Explainable AI improves model performance: By using SHAP (a technique that explains which features matter most), the researchers reduced the number of input variables while actually improving accuracy—test-set AUC increased from 0.9248 to 0.9731. This demonstrates that understanding why a model makes decisions can lead to better predictions, not just transparency.

Automated systems can learn to pick better AI models: The framework uses reinforcement learning (a technique where systems learn through trial and error) to automatically select which machine learning model works best with which features. This reduces the need for human experts to manually test countless combinations, making AI development more efficient and potentially more objective.

Fewer features can mean better results: The study found that using carefully selected subsets of features (rather than all available data) produced more stable and accurate predictions. This suggests that more data isn't always better—strategic selection of relevant information matters for responsible AI development.

Stability matters as much as accuracy: The framework tracked not just whether predictions were correct, but also how consistently accurate they were across different scenarios. This focus on stability is important for real-world applications like 3D printing, where unpredictable performance could cause problems.