The Impact of Social Media Advertising on Consumers’ Purchase Intention: A Study Based on the LSTM Model

Based on 252 consumers' experimental data, this paper examines the relationship between social media advertising information attributes and consumers' purchase intention, and analyzes the mediating effect of perceived value and the moderating effect of brand familiarity. This study proposes to use Long Short Term Memory (LSTM) model to predict consumers’ purchase intension using social media advertising impact, including functionality, and emotionality metrics. Along with these metrics, this study used perceived value as mediating variable and purchase intention as a target variable. The results show that both the functionality and emotionality of social media advertising information attributes have a significant positive effect on consumers' purchase intention, perceived value plays mediating role in the relationship between them, and brand familiarity positively moderates the relationship between emotionality of social media advertising information attributes and consumers' perceived value, that is, there exists a moderated mediation. However, brand familiarity has no moderating effect on the relationship between functionality of social media advertising information attributes and consumer's perceived value. LSTM model achieved 90% training accuracy and 85% testing accuracy, indicating effective LSTM performance for predicting purchase intension of users. The predictive model such as gated recurrent unit (GRU) could be used in future works for better performance results.