Smote-Lasso Model of Business Recovery over Time: Case Study of the 2011 Tohoku Earthquake

A methodology is presented to combine the synthetic minority oversampling technique and the least absolute shrinkage and selection operator to analyze survey data and identify business characteristics correlated with recovery within selected time windows. The methodology addresses challenges that arise when data is imbalanced and predictors are collinear. A case study using data from a survey of business recovery conducted one year after the 2011 Tohoku Earthquake is presented to demonstrate the methodology's application. The survey collected data on 30 predictors describing the physical damage and utility disruptions experienced by the businesses and their sector, size, disaster preparedness, and recovery financing alternatives. The methodology identifies a strong correlation between physical damage and business recovery within 30 days. Industry sector, size, disaster preparedness, and disaster financing become statistically significant when recovery over longer periods is considered. (c) 2021 American Society of Civil Engineers.

Automated summary

The methodology combines synthetic minority oversampling technique with the least absolute shrinkage and selection operator to analyze survey data and identify business characteristics correlated with recovery within selected time windows. It addresses challenges of imbalanced data and collinear predictors, demonstrating strong correlation between physical damage and business recovery within 30 days. Industry sector, size, disaster preparedness, and disaster financing become statistically significant factors when recovery over longer periods is considered.