This study implements the ensemble-based machine learning (ML) approach to predict the temperature-induced properties of high-temperature shape memory alloys (HTSMAs), such as mean martensitic transformation (Mm), mean austenitic transformation ( \({\text{A}}_{{\rm m}}\) ), and thermal hysteresis ( \({\text{T}}_{{\rm th}}\) ), specifically Ti-based HTSMA alloyed with Hf, Zr, Pt, and Pd. Experimentally optimizing the composition for HTSMA during alloy design can be complex, based on the method and economy of alloy making. Three ensemble models were created with different input features, namely, M1 with direct elemental composition as input, M2, and M3, with augmented material descriptors (calculated based on composition) as input. M1 performed better than other models by capturing compositional variation. However, the performance of M2 and M3 is slightly lower and can be a good candidate for understanding the underlying mechanism behind phase transformation behaviour. Critical descriptors based on structural, thermodynamic, and electronic parameters were found to affect transformation behavior, which is correlated with elements considered in alloying in NiTi and Ti-based HTSMAs. The current study shows the effectiveness of ensemble models in predicting temperature-induced properties and gives insights into the effect of thermodynamic, structural, and electronic properties on the phase transformation behavior of HTSMA, which could provide valuable inputs during alloy design.
