Emotions and machine learning in innovation within Mexico’s sustainable used-vehicle market

Abstract

Context. The used-vehicle market in Mexico is essential for mobility and the circular economy. However, traditional studies prioritize rational variables, overlooking emotional factors.

Problem. Traditional linear models fail to capture the interdependence of emotional responses. Thus, the Research Question arises: To what extent do emotions predict loyalty in Mexico’s used-vehicle market?

Objective. To evaluate the impact of emotions on consumer loyalty using machine learning techniques to optimize decision-making in this sector.

Methodology. A quantitative study was conducted with 1,000 buyers in Aguascalientes. The PANAS instrument was used, and a logistic regression model was compared against a Random Forest model. The superiority of the non-linear model was validated using the DeLong test (Z = 3.84; p < 0.001).

Findings. The Random Forest algorithm achieved 87% accuracy. Satisfaction and security are the main predictors of loyalty, while fear and confusion act as critical purchasing barriers.

Originality. The study provides a theoretical advancement by integrating neuroeconomics with data science (Scientia). Additionally, it offers a predictive tool for companies in the sector to design sales strategies based on the customer's emotional experience (Praxis). It provides a framework for reducing emotional barriers in the second-hand market and contributes directly to SDG12.

Conclusions and limitations. The findings confirm that emotions are robust predictors of pre-owned vehicle purchase decisions and that their strategic management is critical for fostering responsible consumption. The primary limitation lies in the use of a non-probabilistic sample from a single region. Future research should employ longitudinal designs and replicate the study across diverse contexts while maintaining the multidisciplinary approach.