Multi-walled carbon nanotubes (MWCNTs) enable efficient photothermal coatings for thermal management applications due to their broadband near-infrared absorption and high thermal conductivity. This study optimizes MWCNT-polymer formulations using random forest machine learning (ML) to maximize photothermal conversion efficiency (η), targeting inputs like MWCNT concentration (1–10 \({\text{mass}}\%\) ), polymer type (polystyrene, polyethylene, and polyurethane), and coating thickness (100–500 nm) prepared via dip, spray, or spin coating on glass substrates. A dataset of 500 + experimental points (features: composition, processing parameters; target: η and steady-state temperature rise) underwent preprocessing (normalization and categorical encoding) and fivefold cross-validation. The random forest model achieved R2 = 0.93 (validation), outperforming baselines by predicting optimal formulations (e.g., 5 \({\text{mass}}\%\) MWCNT in polyurethane yielding η = 85%, ΔT = 35 °C under 1 sun irradiation in 10 min). ML-guided coatings showed uniform dispersion (SEM-confirmed), mechanical robustness, and stability over 20 cycles. Compared to MXene/Au hybrids, MWCNTs offer cost-effective scalability despite agglomeration risks, mitigated here via optimized dispersion. This work demonstrates ML-accelerated design of high-performance photothermal coatings, with future extensions to hybrid systems and real-time prediction.
