In terms of the improvement of efficiency and data maintenance of the ChemiCloud platform, we added the public page of environmental distribution survey results. Six internal query functions and two customized functions were developed. We also positioned the coordinates and verified 3,148 factories whose coordinates were not provided. Additionally, we expanded three data systems, provided 3 APIs to fire departments, and continued to maintain fifty-three chemical substance management information systems. A total of one hundred and forty-eight ChemiCloud accounts were created. We also continued to carry out platform information security, operation and maintenance. Last but not least, we completed five employee trainings, thirteen meetings with authorities concerned and one cross-functional meeting.
In terms of platform improvement, we completed the inventory of front-end and back-end programs, data tables, and indexes of the ChemiCloud platform. Furthermore, we analyzed the system’s function performance, database query speed, system log and developed mid-term and long-term system optimization plans. In addition, one of the cases is used to demonstrate the developed solutions for adjusting the data integration architecture and improving application flexibility. Last but not least, a site map, one geographic map optimization and 3 function related optimizations were completed.
In terms of big data analysis, we completed the one-time login (OTP) service connection with other systems and manufacturers. We also completed the open authorization (OAuth) data connection, and demonstrated record of 1,2-dichloropropane and (4-Ethoxyphenyl)urea on the chain. In addition, we completed the development and testing of blockchain wallets, transaction vouchers and chemical substances pass functions. We also carried out the actual demonstration of blockchain system concept verification. In the part of the high-risk abnormal manufacturer detection model, Variational Graph AutoEncoder (VGAE) is used to extract the interaction characteristics between manufacturers and chemical substances. Self-Supervised Learning (SSL) is used to detect and perform end-to-end training. In the news crawler part, different technologies are used for foreign information translation, Chinese and English chemical substance names and food names are used to improve the recognition level. In addition, interviews are conducted with the Taiwan Food and Drug Administration for the pilot program of value-added promotion. We selected 6 foreign news and 5 domestic news for the Taiwan Food and Drug Administration, asking for its help and feedback.
In terms of other analytical works, the environmental correlation analysis of Erren River and Wu River were completed, and a list of 1,023 groups of 2,837 chemicals with the same function and a system query function were established. We reviewed and updated relevant domestic and foreign regulations list of chemical substances of concern regularly, and the number of chemical substances in the list has increased from 32,830 to 33,638. In addition, the collection of 2022 domestic food-related inspection and sampling results was completed.