DEVELOPING A MACHINE LEARNING MODEL UTILIZING THE SUPPORT VECTOR MACHINES (SVM) TECHNIQUE

ABSTRACT

Titanium Nickel Stannide (TiNiSn), a half-Heusler alloy, demonstrates a high electrical conductivity, which means it is good at carrying electrical current. Previous studies have provided evidence to support the assertion that TiNiSn has shown several desirable characteristics, notably a significantly elevated bulk modulus. This characteristic renders it a potential candidate for various applications, encompassing the fields of structural engineering and thermoelectricity. In this study, we provide two methodologies, namely first principle calculations, to analyze the bulk modulus of TiNiSn. The obtained value of around 134.6 GPA for the bulk modulus indicates the material’s robustness and stiffness. The aforementioned feature assumes a pivotal role in its various applications within the fields of spintronics and thermoelectricity. In this study, a machine learning (ML) model is employed to forecast the bulk modulus. A total of 2382 ternary compounds with elastic data were acquired from the Aflow datasets through the utilization of the Application Program Interface (API). The estimation of the Root Mean Squared Error (RMSE) was associated with a Standard Error of 17.09. Additionally, the Uncertainty Calibration was measured at 0.919, indicating a calibration uncertainty percentage error of 2.6%. The models that have been built exhibit a significant level of promise in terms of their accuracy, precision, and ease of implementation, which enhances their practical usefulness.