Classification and Prediction of Food Safety Policy Tools in China Based on Machine Learning

Abstract

Governments use policy interventions to mitigate food safety risks. Despite its crucial role, empirical studies evaluating the effectiveness of China's food safety policy tools are scarce. Drawing on a dataset encompassing 11,236 food safety policy texts from 2005 to 2021 and the incidence of problematic food products in the Eastern, Central, and Western regions of China, this study employs Latent Dirichlet Allocation (LDA) and eXtreme Gradient Boosting (XGBoost) models to facilitate the classification of policy tools and forecast the effectiveness of policy combinations. The study reveals that (1) local governments have gradually become an important supplementary maker of food safety policies, and have issued an increasing number of policy tools year by year. (2) Mandatory policy tools are predominant in number and have the highest legal hierarchy and authority levels, followed successively by guiding policy and voluntary policy tools. (3) Mandatory policy tools demonstrated the most effective intervention results, followed successively by guiding policy and voluntary policy tools. (4) The forecast analysis reveals that combinations of policies within high-growth frameworks and those driven by mandatory regulations emerge as the most effective. Therefore, the balance of policy tools in terms of type, effectiveness, and quantity, as well as their applicability in different situations, should all be taken into account.

Keywords

• eXtreme Gradient Boosting
• Food safety policy
• Food safety risk
• Policy effectiveness
• Policy tools