Restaurant survival prediction using customer-generated content: An aspect-based sentiment analysis of online reviews

Business failure prediction or survival analysis can assist corporate organizations in better understanding their performance and improving decision making. Based on aspect-based sentiment analysis (ABSA), this study investigates the effect of customer-generated content (i.e., online reviews) in predicting restaurant survival using datasets for restaurants in two world famous tourism destinations in the United States. ABSA divides the overall review sentiment of each online review into five categories, namely location, tastiness, price, service, and atmosphere. By employing the machine learning-based conditional survival forest model, empirical results show that compared with overall review sentiment, aspect-based sentiment for various factors can improve the prediction performance of restaurant survival. Based on feature importance analysis, this study also highlights the effects of different types of aspect sentiment on restaurant survival prediction to identify which features of online reviews are optimal indicators of restaurant survival.

Automated summary

This study investigates the impact of customer-generated content, specifically online reviews, on predicting restaurant survival using aspect-based sentiment analysis. The results show that aspect-based sentiment, focusing on specific factors like location, taste, price, service, and atmosphere, improves the accuracy of restaurant survival prediction compared to overall review sentiment. Additionally, the analysis of feature importance identifies which aspects of online reviews serve as optimal indicators for restaurant survival.