Prediction of Chemical Reactivity Parameters via Data‐Driven Approach

Abstract

Novel material designing in an efficient way and its property prediction is empowered by data‐driven approach. For system designing or synthesis, stable and compatible chemical counterparts containing functional materials are preferred. In this regard, the knowledge of chemical reactivity is indispensable and is closely associated with how a substance reacts in a particular chemical reaction. In this work, chemical reactivity parameters of some organic molecules through machine learning (ML) algorithms are predicted. Several categories of descriptors are used as input features to predict HOMO‐LUMO energy gap, ionization potential, electron affinity, chemical potential, chemical hardness and electrophilicity index. The accurately achieved reactivity parameters confirm the descent training of the model from the integrated data of organic molecules. This work confirms that chemical properties reproduced through ML approach are closely correlated with density functional theory (DFT) ‐based results, so the proposed ML approach provides reactivity information at a very cheap cost. The prediction of chemical reactivity, as well as perception of the correlations between input features and targeted properties of organic molecules, may lead the experimentalist to know more about the observation.