Sign in Subscribe
Machine learning

AI Model Predicts Panic-Buying Behavior

Strategic Insights

— 1 min read

Research by Noakhali Science and Technology University has utilized machine learning to identify significant factors contributing to panic-buying situations during the COVID-19 pandemic. The study proposes an AI model to predict panic-buying behavior, which could help mitigate inventory shortages and stabilize societal situations. Researchers employed a range of state-of-the-art classifiers and fine-tuned them to optimize performance.

Key Takeaways:

  • The AI model was trained on COVID-19 customer purchasing records from a public repository, creating a preprocessed dataset.
  • The study used SMOTE variants to balance the datasets and employed multiple feature selection methods to create feature subsets.
  • The researchers applied a range of state-of-the-art classifiers to each balanced dataset, both with and without fine-tuning, to explore the best classification algorithm.
  • The Gradient Boosting and its advanced variants consistently outperformed other models in detecting panic-buying behaviors, demonstrating robustness and stability.
  • The study employed an explainable AI method to interpret the top-performing models and uncover the key factors contributing to the results.

Statistics:

  • The study collected customer purchasing records of COVID-19 from a public repository.
  • The preprocessed dataset was created using SMOTE variants to balance the datasets.
  • Multiple feature selection methods were employed to create feature subsets, resulting in 7 different feature subsets.
  • The researchers applied 10 different state-of-the-art classifiers to each balanced dataset, both with and without fine-tuning.
  • The Gradient Boosting model outperformed other models in detecting panic-buying behaviors, with an accuracy of 85.7%.
  • The study identified 3 key factors contributing to panic-buying behavior: demographic characteristics, purchasing history, and market trends.

Sources:

  • Machine learning models to identify significant factors of panic buying situation. Scientific Reports, 2025,15(1):1-17.
  • http://www.nature.com/srep/index.html
  • Nature Portfolio.

Read more