資料庫系統輸入資料時，可能有資料填寫錯誤的情況發生，導致統計或資料分析產生不正確之結果。以臺灣地區橋梁管理資訊系統(Taiwan Bridge Management System，TBMS)為例，該系統之資料由使用者透過網路或行動裝置輸入，使用者可能因量測錯誤或判斷錯誤，甚至輸入時之誤填導致基本資料欄位內容錯誤、基本資料欄位間之資料互相衝突或檢測結果之資料不合理。雖近年來雖有部分現地查核制度以提升基本資料與檢測資料之正確率，但因全國橋梁總數高達28,000座，無法全面自動化查對資料之正確性。為提升整體系統資料之正確性，本計畫擬針對臺灣地區橋梁管理資訊系統之基本資料及檢測資料建立智慧型之自動化︰（1）在基本資料方面，針對欄位內容及欄位間互相衝突之錯誤資料，以專家訪談結果整合各欄位數值範圍及合理的橋梁設計規則，建立基本資料自動化偵錯之模式；（2）在檢測資料方面，本研究擬利用K-means結合PSO(Particle Swarm Optimization)，利用基本資料40個關鍵因子，將相似屬性之橋群做分類分群，利用分析結果找出相似橋群間劣化構件之規則，自動化判斷出明顯填寫錯誤之檢測數值以供修正。本計畫執行完成後，將能對台灣地區之橋梁管理與維修經費之正確運用提供相當之助益。 ;Error Data inputting into a database results in incorrect statistic or analytic results based on the database. In the Taiwan Bridge Management System (TBMS), inventory and inspection data are input through network browsers or portable devices. However, those data could be mistaken due to incorrect measurements, judgments, or even typos. Typical data errors include incorrect basic data, conflicts among basic data, and unreasonable inspection results. Recent remedy actions such as checking at bridge sites only obtained partial not global success since there are more than 28,000 bridges in the database of TBMS.The main objective of this research is to establish automated error data detection mechanisms in the TBMS to improve data correctness of the system. The mechanisms include (1) detection of error basic data, in which data limitations and data rules among various fields will be obtained and summarized through expert interviews; (2) detection of error inspection data, in which bridges will be firstly categorized by the K-means and Particle Swarm Optimization (PSO) using 40 bridge basic data, then suspicious wrong inspection data will be prompted if they are materially different from that of similar bridges. Results of this research will be very helpful and beneficial for the bridge management agencies that use TBMS for management and distributing maintenance budgets of the bridges.
