Machine Learning Assisted Designing of Dyes and Oscillator Strength Prediction: Chemical Space Generation and Visualization

Abstract This study introduces a swift framework utilizing machine learning (ML) techniques to design dyes. The framework involves training of ML models to predict oscillator strength. Subsequently, employing the breaking retrosynthetically interesting chemical substructures (BRICS) methodology, 10,000 new dyes are generated, and their oscillator strength values are forecasted using a pre‐trained ML model. Dyes with higher predicted oscillator strength values are retained for further consideration. To make future experiments easier, the synthetic accessibility of these chosen dyes is also assessed. The structural diversity among the chosen dyes is investigated, revealing a high degree of variation.