A Transformer-LSTM-SVR hybrid model for AI-driven emotional optimization in NEV embedded interior systems.

With the rapid expansion of the new energy vehicle (NEV) market, a shift from function-oriented toward emotion-oriented interior design has emerged as a key factor in enhancing user experience and brand differentiation. Current research on emotion optimization in new energy vehicle interiors lacks efficient nonlinear modeling methods. To address this gap, a hybrid Transformer-LSTM-SVR model is proposed, which significantly enhances emotion prediction accuracy by integrating attention mechanisms and temporal modeling. This model addresses the complex nonlinear relationship between users' emotional satisfaction and interior design attributes. Furthermore, interpretability analysis reveals key design feature differences among user groups. The proposed method combines a Transformer module, which captures higher-order interactions among multidimensional design parameters (e.g., sentiment evaluation coefficients and task completion time), and a Long Short-Term Memory (LSTM) network, configured to enhance time-series feature capture through adjustments to hidden unit count and sequence length. Potential representations from both modules are combined into high-dimensional vectors via a feature fusion mechanism and subsequently fed into a Support Vector Regression (SVR) module for fitting nonlinear relationships. This hierarchical architecture effectively mitigates the limitations of traditional SVR in modeling dynamic time series and nonlinear relationships, while enhancing model robustness through the synergy between global context-awareness and time-series dependency. Performance comparisons against benchmark models-including traditional SVR, particle swarm optimization (PSO)-tuned SVR, PSO-tuned random forest (RF), Backpropagation Neural Network (BPNN), and Gradient Boosting Regression (GBR)-demonstrate that the proposed model significantly outperforms these baselines, improving prediction accuracy by 12.7% to 23.4%. Compared to traditional KE methods, the synergistic integration of the LSTM's temporal attention mechanism and the Transformer's global context modeling improves system robustness against noisy user feedback data by 18.9%. Results indicate that the model significantly enhances the prediction accuracy of users' emotional needs, offering a viable approach for emotion-oriented and sustainable interior design in new energy vehicles. This study provides practical tools for designers, thereby enhancing the market competitiveness of new energy vehicles and promoting sustainable development.