Data-driven styling optimization of an intelligent medical waste recycling robot using AHP-CRITIC and SSA-ELM

Abstract To improve the attractiveness and practicality of intelligent medical waste recycling robots, this study proposes a data-driven styling optimization framework integrating Kansei Engineering, AHP-CRITIC weighting, and a Sparrow Search Algorithm-optimized Extreme Learning Machine (SSA-ELM). Factor analysis was first used to reduce affective vocabularies and extract core perceptual dimensions. AHP-CRITIC was then applied to identify five key styling features closely associated with user perception: side profile, front view, display screen, wheels, and body. Based on these features, SSA-ELM was employed to establish a nonlinear mapping between design variables and perceptual responses, and to predict the optimized styling combination. Model performance was evaluated using repeated 10-fold cross-validation. The results show that SSA-ELM generally outperformed GA-ELM, PSO-ELM, and conventional ELM in predictive accuracy and stability, with the most robust performance observed in the Balanced dimension. User evaluation further confirmed that the optimized design significantly improved ratings of balance, smoothness, perceived intelligence, reduced work burden, and safety. The proposed framework provides a quantitative and interpretable pathway for the affective design of intelligent medical service equipment.