Unveiling the development trends of environmental and human health concerns for pesticides: Integrating an intelligent approach and data mining across diverse databases

The impact of pesticide use on environmental and human health has been a persistent global concern. In the era of big data, the scientific literature concerning big data is a significant source of information; however, it is difficult to construct an optimal policy based on traditional insight using keyword searches or a single static-specialized database. In this study, we constructed a new path for data mining across multiple databases to provide a comprehensive picture of the major issues concerning environmental pollution and human health as a result of pesticide use at the global scale. This approach uses a classic unsupervised learning algorithm, Latent Dirichlet Allocation (LDA), in combination with a newly developed dataset of pesticide-associated human health outcomes (PAHHO), including 618 health outcomes classified into 14 types of toxic effects. Our data visualization revealed a shift in the scientific center for pesticide research over the past five decades. The major issues concerning environmental pollutants and health outcomes varied among different countries and in different periods, which was verified in our analysis of several organochlorine pesticides (OCPs) about which people are particularly concerned. A cooccurrence network of adverse health outcomes has gradually increased, suggesting that the impact of pesticides on human health is persistent and cumulative. Our work not only provides a promising research direction related to the most concerning issues in a systematic and visualized way but also provides valuable references to formulate optimal strategies for the goal of the global "One Health" objective in pesticide regulation.