按 Enter 到主內容區

108年跨部會化學物質資訊服務平台 (化學雲)應用計畫
To Apply the Chemical Substances Management Information Project

  • 發布單位:毒物及化學物質局
108年跨部會化學物質資訊服務平台 (化學雲)應用計畫
專案計畫編號
經費年度2019
計畫經費 千元
專案日期2019-02-20 ~ 2019-12-15
專案主持人賴辰瑜
主辦單位毒物及化學物質局
承辦人林桂如
執行單位關貿網路股份有限公司
專案分類其他
中文關鍵字跨部會化學物質、機關跨域合作、警示和快報
英文關鍵字Inter-Ministerial Chemical Substance, Inter-Ministerial Cooperation, Alert and Quick Report
協同主持人許震洋
共同主持人
計畫聯絡人姚聖屏
中文摘要

資料擴增及維護方面,本年度維持46個系統持續拋轉,開立994個化學雲帳號,稽催5個系統拋轉,並擴增介接6個資料來源;開發廠商資料整併程式,整併化學物質計10,723筆,廠商資料整併表計14,005筆。透過資料清洗及正規化處理,提升化學物質及廠商資料鏈結度,化學物質及廠商名稱可正規化程度分別提升至69%及85%。部會需求開發及介接方面,共計開發及精進12項主要功能,包含:進行網頁改版、建置5個應用主題專區、簡化手持裝置瀏覽效果、開發化學物質運作資訊公開、整合防災資訊擴增2項資料來源、提供毒性化學物質許可管理系統申報資料及工廠危險物品申報網資料2項Web Service功能供部會介接使用。大數據分析方面,本計畫導入巨量資料分析技術及人工智慧技術,利用化學雲整合各部會既有資料及外部資料,已完成4項大數據分析功能開發;在業者風險分析部分,以貝氏網絡分析方法建置預測模型,篩選風險廠商提供業務稽查單位參考,以降低稽查人力、時間與金錢成本;在化學物質業者關聯分析中,結合化學雲與電子發票交易資料,運用社會網絡分析,對食安辦建議37項優先列管及經濟部中辦57項具食安風險疑慮化學物質進行交易流向分析,當交易網絡所含廠商數量龐大,實際上難以迅速透過視覺化掌握交易關係,利用社會網絡方法,計算各項中心性分數,瞭解業者交易關係,或何業者處於守門人的角色,廠商間聯絡須透過該家廠商以達成交易,輔助在複雜網絡中,快速掌握各類關鍵廠商。針對國內外化學物質相關新聞及網站,自2018年5月至今,已擷取國內新聞約10,121則及國外警訊5,743則進行分析,透過自動翻譯程式及文字探勘技術,辨識新聞屬性及主題,掌握時事議題之核心。在分階段推動電子發票之可行性方面,從法規、業者端及管理效益三個面向進行評估及規劃,提出可能遭遇之問題及解決方案,並以財稅資料應用於毒化物管理提出應用情境,供化學雲進行毒物及化學物質源頭管理、流向追蹤及勾稽查核等決策參考。在化學物質單一服務窗口部分,分析各部會法規,比對各系統重疊化學物質及欄位,提出9個建議試行機關系統,融合委員意見提出兩個執行方式,分別為申報資訊整合推送以及單一服務介面,作為未來執行之參考。危險物品運作廠家警示因子分析部分,已完成603個危險物品與近7,000家廠商警示分析因子之資料收集;依分析因子分別擬出危險物品與廠家的警示分數計算方式,產出危險物品與廠家警示優先順序清單,供消防單位參考使用。

英文摘要

In terms of data amplification and maintenance, in this year, we’ve maintained 46 systems under continuous running, registered 994 ChemiCloud accounts, monitored 5 systems running, and expanded access to 6 data sources. We’ve developed vendor’s data integration programs and integrated 10,723 chemical substances and 14,005 vendor data. Through data cleaning, information logic and manual judgment of normalization, the available normalizing grade of chemical substances and vendor names can be upgraded to 69% and 85%, respectively. In terms of the development and interface of department requirements, we’ve totally developed and refined 12 main functions, including: web page revision, built 5 application theme zones, simplified the browsing effect of handheld devices, developed the disclosure of operating information on chemical substances, integrated disaster prevention information and expanded two data sources, providing notification materials for the toxic chemical substance permission management system and factory hazardous materials declaration network information, two web service functions are applicable for department interface. In terms of big data analysis, this plan introduces data analysis technology on mega data and AI technology; use ChemiCloud to integrate the existing data and external data of various departments; and has completes the development and construction of four big data analysis functions. In vendor risk analysis, the Bayesian network analysis method is used to build a prediction model, provide a list of referenced risky vendors for the business audit entity to reduce the audit labor, time and expenses. In the chemical substance industry association analysis, combine ChemiCloud and electronics invoice transaction data; use social network analysis method, carried out the food safety operation; made 37 suggestions on food safety in-priority management and conduct 57 food safety risk suspected chemical substances for transaction flow analysis with the Middle Office of MOEA. When the vendor amount involved in the trading network is enormous, it is practically difficult to quickly master the trading relationship through visualization; it must use social networks to calculate the centralities, realize vendor’s transaction relationship, or which vendor is taking the gatekeeper role, communications among vendors are achieved through that gatekeeper vendor to complete the transaction; which helps to quickly master various key vendors in a complex network. For related domestic and foreign chemical substances news and websites, since May 2018, about 10,121 domestic news and 5,743 foreign warnings have been collected for analysis. We use text exploration and document topics classification to identify the attributes and topics of news and to master the core of current affairs issues. Regarding the feasibility of promoting electronic invoices in stages, we evaluate and plan for the tasks from the aspects of laws / regulations, vendors and management effectiveness; propose possible problems and respective solutions; propose application scenarios based on fiscal and taxation data for toxicant management, and provide a reference for decision making of ChemiCloud for toxicant and chemical source management, flow tracking, and cross-checking. In the single service window of chemical substances, we analyzed the laws / regulations of various departments, compare the overlapping chemical substances and columns of each system, set forth 9 suggested trial agency systems; and proposed two implementation methods per the committee members’ opinions, which are the integration of application information and the single service interface as a reference for future implementation. At present, the analysis of warning factors for dangerous articles operating vendors has now completed the collection of warning analysis factors for 603 dangerous articles and nearly 7,000 vendors. According to the analysis factors, the calculation method of warning scores for dangerous articles and vendors has been drawn up. A priority list of dangerous articles and vendors' warnings is produced for the firefighting entities to refer thereof.

附件下載

檔案名稱 檔案大小 下載次數
9162104211800003e3.pdf 14 MB 409
分享
  • facebook(另開新視窗)
  • line(另開新視窗)
  • 列印
  • 點閱次數:211
  • 資料發布日期:110-04-21
  • 資料更新日期:110-04-21
  • 資料檢視日期:111-08-12
回上一頁回頁首