Random forest regressor for predicting sensory texture of emotional designed packaging films

• A random forest regression model was developed to predict sensory texture preferences of packaging films. • Surface roughness parameters (Ra, Ry, Rz, Rq, R-MAD) were measured using ISO-compliant methods. • Emotional design elements were evaluated to enhance consumer appeal for packaging films. • Textural attributes of matte and varnish-textured films were linked to sensory feedback. • The proposed model demonstrates high accuracy and practicality for industrial applications. A random forest (RF) regression model was developed to predict sensory texture preferences of packaging films, enhancing their emotional appeal to consumers. Five films, including matte and varnish-textured prints, were analyzed using a surface profilometer to measure roughness parameters ( R a, R y, R z, R q, and R -MAD) in compliance with ISO 4287 and ISO 24118–1. Sensory preferences were evaluated through tests involving 75 panelists, and correlations between roughness parameters and preferences were established. Power spectral density (PSD) analysis with Welch window preprocessing provided detailed surface texture insights. The RF model achieved a coefficient of determination of 0.977, outperforming partial least squares regression, and highlighted the importance of significant wavelength regions in predictive modeling. This study demonstrates a robust framework for integrating machine learning in packaging design to optimize sensory appeal.