Data‐Driven Molecular Architectonics for Dye Discovery: Integrating Gradient Boosting Paradigms With Retrosynthetic Fragmentation for Band Gap Optimization

The current research is based on data‐driven chemical space generation of dyes and prediction of their band gaps by employing machine learning techniques. A wide range of dyes is used in the current research, along with various machine learning techniques, to check which one is best suited for the accurate prediction of band gaps. The results obtained in the current research clearly prove the effectiveness of the Light Gradient Boosting Machine Regressor by achieving high accuracy with minimal errors in prediction. The chemical space of 10,000 new dyes is generated and then used for screening by predicting their band gaps and checking their desirable electronic properties. Moreover, a synthetic accessibility analysis is conducted, proving that these dyes can be easily synthesized.