Data-driven thermoelectric modeling: Current challenges and prospects

Recent advancements in computing technologies coupled with the need to make sense of large amounts raw data have renewed much interest data-driven materials design and discovery. Traditional science research relies heavily on experimental gauge properties materials. However, this paradigm is purely based trial error ongoing can take decades discover new Data-driven modeling tools such as machine learning its proven libraries help speed up materials’ discovery process through implementation powerful algorithms readily available material datasets mined from ever-increasing private- government-funded databases. In Perspective, we applied various models tens hundreds thermoelectric compounds obtained density functional theory calculation results. our preliminary analysis, made use pymatgen library matminer add explore key features that propensity accurately predict achievable target output. We evaluated accuracy performance coefficient determination (R2), root mean square error, K-fold cross-validation metrics identified most important descriptors for Finally, reviewed current state-of-the-art discovery, challenges, prospects.