Interpretable Machine Learning Model on Thermal Conductivity Using Publicly Available Datasets and Our Internal Lab Dataset
Bibliographic record
Abstract
Machine learning (ML), a subdiscipline of artificial intelligence studies, has gained importance in predicting or suggesting efficient thermoelectric materials. Previous ML studies have used different literature sources or density functional theory calculations as input. In this work, we develop a ML pipeline trained with multivariable inputs on a massive public dataset of ∼200,000 data utilizing a high-performance computing cluster to predict the thermal conductivity (κ) using four test sets: three publicly available datasets and a dataset built using previously published data from our own group. By taking advantage of this massive dataset, our model presents an opportunity to further expand the understanding of the selection of features with various thermoelectric materials. Among the several supervised ML models implemented, the eXtreme Gradient Boosting algorithm (XGBoost) turned out to be the best method during the 5-fold cross-validation method, with their averaged evaluation coefficients of R 2 = 0.96, root mean squared error ( RMSE ) = 0.38 W m −1 K −1, and mean absolute error ( MAE ) = 0.23 W m −1 K −1 . Additionally, with the aid of feature selection and importance analysis, useful chemical features were chosen that ultimately led to reasonably good accuracy in the series of test sets measured as per the evaluation coefficients of R 2, RMSE, and MAE, with values ranging from 0.72 to 0.89, 0.52 to 1.08, and 0.40 to 0.66 W m −1 K −1, respectively. Checking the worst outliers led to the discovery of some errors in the literature. Postmodel prediction, the SHapley Additive exPlanations (SHAP) algorithm was implemented on the XGBoost model to analyze the features that were the key drivers for the model’s decisions. Overall, the developed interpretable methodology produces the prediction of κ of a large variety of materials through the influence of chemical and physical property features. The conclusions drawn apply to the research and applications of thermoelectric and heat insulation materials.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".