Prediction of electrical conductivity, Seebeck coefficient, thermal conductivity, and ZT for p-type SnSe with Na 1% doping
1) Alloying element and ratio
2) Sn vacancy ratio
ML
NIMS ZT Profile Prediction for SnSe Materials
Prediction the figure of merit ZT profile in Temperature (multi-target)
1) Host element composition & ratio
2) Dopant element composition & ratio
3) Dopant element properties
Machine Learning Prediction Model for the Temperature Profile of ZT
1. Thermoelectrical Material Experimental Data ans features
We are using the TEXplorer data set, collected from the real experiments. In feature modeling element property features are added in data set, especially for dopant elements. Also, the compositions and ratios of host and dopant elements are included in feature set.
2. Machine Learning Prediction Model
A few machine learning algorithms provide a capability to predict multi-target, here, the temperature profile of ZT. We adopt the random forest algorithm for it.
3. Results
We compared the performance between single target-multiple prediction and multiple target-single prediction, which is implemented in TEXplorer. Multiple target-single prediction with random forest gives better results in profile predictions.
ML
Predicting toolkit of thermoelectric properties for Bi2Te3 materials
Prediction of electrical conductivity, Seebeck coefficient, thermal conductivity, and ZT for Bi2Te3 with various dopants
