This study facilitates the data-driven design of novel cast TiAl alloys by systematically investigating the critical elements and their interactions affecting room-temperature (RT) tensile properties by the machine learning method based on SHAP analysis. Comparative analysis of three algorithms within the training dataset proved the random forest regression (RFR) as the optimal modeling approach. To evaluate model performance and prevent overfitting, leave-one-out cross-validation (LOOCV) was simultaneously implemented during training. All the three well-trained models demonstrated robust predictive capabilities for ultimate tensile strength (UTS), elongation (EL), and yield strength (YS). Detailed investigation on both the magnitude and directionality of feature importance and interaction disclosed distinct elemental influences: B, C, and Nb predominantly improved UTS and YS, while Cr, Mn, and Al positively affected EL. The highly probable direction of feature interaction between two different elements on the RT tensile properties of cast TiAl alloys was basically revealed. Notably, Al–B interactions enhance UTS at Al < 45.5 at%; Cr–Mn synergistically improves EL when Cr > 1 at%; both Al–B and Al–C interactions boost YS within 44–46 at% Al. Despite a slight distinction in casting technology, this research established a qualitative relationship between the chemical elements and the RT tensile properties of TiAl alloys, providing design recommendations for cast TiAl alloys with excellent RT tensile properties.

Hypergraphs are powerful tools for modeling complex systems because they naturally encode higher-order interactions. However, most existing hypergraph representation-learning methods still struggle to capture such high-order structures, particularly in heterogeneous hypergraphs, which results in suboptimal performance on structure-sensitive tasks such as node classification. This paper presents WCRW-MLP, a new framework that integrates a Weighted- and Clustering-Biased Random Walk (WCRW) with a multi-layer perceptron. WCRW extends second-order random walks by introducing node-pair co-occurrence weights and triadic-closure clustering bias, enabling the walk to favor structurally significant and locally cohesive regions of the hypergraph. The resulting walk sequences are processed with Skip-gram to obtain high-quality structural embeddings, which are then concatenated with node attributes and fed into an MLP for classification. Experiments on several real-world hypergraph benchmarks show that WCRW-MLP consistently surpasses state-of-the-art baselines, validating both the efficacy of the proposed biasing strategy and the overall framework. These results demonstrate that explicitly modeling co-occurrence strength and local clustering is crucial for effective hypergraph embedding.