Very important review on “Quantitative StructureProperty Relationship Modeling of Diverse Materials Properties” has came out from Australian group in Chemical Reviews.
“Only one report of QSPR modeling of electrocatalysts has been published. The electrochemical performance of six samples of nonplatinum porphyrin-based catalysts of oxygen reduction was predicted based on 24 XPS spectral variables and electrochemical measurements. The combination of genetic algorithm and multiple linear regression generated a model that had excellent predictivity for the training set and good cross-validation performance. However, the imbalance between the small data set size and number of descriptors risks overfitting the QSPR model.”
Note to myself: make students burn more samples!