利用大數據建立營建人員安全管理的風險分類模型

A Big Data Based Safety Risk Classification Model of Construction Workers for Construction Site Safety Management

相較於其他行業，營建業因事故發生率高而常受到各界的指責與批評，故施工安全課題一直受到各國政府的高度重視。人工智慧（Artificial Intelligence, AI）的應用，特別是在電腦影像識別的應用，近10年來在營建工程安全自動監控方面取得了成功的應用。然而，常因為工程人員於年齡、工作年資、工種等不同的個人屬性，導致施工事故的發生機率可能會有所不同，這些屬性是現有的人工智慧技術無法識別的。本研究利用勞動部勞動及職業安全衛生研究所建置的「營造業重大職災知識平臺」中積累的大數據來確定營建工程的安全風險級別。根據工人的屬性，如年齡、資歷、工程類型、工程規模、所屬公司規模等對工人進行分類。透過與其他電腦識別技術相結合，本研究提出的施工人員安全風險分級方法，可以根據不同的關注程度來判斷施工人員的危險程度。通過本研究提出的施工安全自動監控系統的應用框架，將有助於現場管理人員將注意力特別集中於可能遇到即時風險的工人身上，如此將有助於減少工地施工事故的發生。

The construction industry has been blamed for her high accident occurrence rate, compared with any other industries. The construction safety has drawn major attention of the government world-wide. The application of Artificial Intelligence (AI), especially in computer image recognition, has enjoyed successful applications for automatic monitoring of construction safety in the past decade. However, the probability to cause construction accidents may differ due to different personal attributes such as age, professional experience, expertise, etc. Such attributes are not distinguishable by existing AI techniques. This study utilizes the big data accumulated in a national database—the Construction Industry Accident Knowledge Platform (CIAKP)—developed by the Institute of Labor, Occupational Safety and Health (ILOSH), Ministry of Labor, Taiwan, to determine the safety risk levels of construction workers based on his/her attributes, e.g., age, professional experience, expertise, project type, cost of project, size of company, etc. By integrating with other intelligent computer visualization techniques, the construction personnel safety risk classification method proposed in this research can pay attention to the construction worker's according to different risk levels. With the application framework of the construction safety automatic monitoring system proposed by this research, it will assist site managers to focus their attention on workers who may encounter real-time risks. As a result, it will help reduce the occurrence of construction accidents on job sites.