Quantitative structure-property relationships of electroluminescent materials: Artificial neural networks and support vector machines to predict electroluminescence of organic molecules

Electroluminescent compounds are extensively used as materials for application in OLED. In order to understand the chemical features related to electroluminescence of such compounds, QSPR study based on neural network model and support vector machine was developed on a series of organic compounds commonly used in OLED development. Radial-basis function-SVM model was able to predict the electroluminescence with good accuracy (R = 0·90). Moreover, RMSE of support vector machine model is approximately half of RMSE observed for artificial neural networks model, which is significant from the point of view of model precision, as the dataset is very small. Thus, support vector machine is a good method to build QSPR models to predict the electroluminescence of materials when applied to small datasets. It was observed that descriptors related to chemical bonding and electronic structure are highly correlated with electroluminescence properties. The obtained results can help in understating the structural features related to the electroluminescence, and supporting the development of new electroluminescent materials.