The Polymer Society of Korea

Abstract

The solution phase behavior of thermosensitive polymers governs material processing, synthesis, and ultimate application performance. However, determining and controlling the cloud-point temperature (T_cp) remains heavily reliant on trial-and-error screening. Here, we establish data-driven machine learning models to predict T_cp for thermosensitive polymer solutions. The models are trained on a compiled dataset that includes molar-mass variables, repeating unit information, and solution parameters such as pH. Each record is standardized and encoded into harmonized, machine-readable descriptors for model input. Predictive performance is evaluated using metrics including mean absolute error (MAE), providing a robust and reusable benchmark for T_cp prediction. This work offers a methodological framework and a performance reference to support subsequent model optimization, mechanistic studies, and materials design in expanded formulation spaces.