民國90年至94年肇事件數有逐年攀升的情形。以民國94年為例，我國肇事件數總計155,814件（平均每日有427件的肇事發生）尚未包含當事人自行和解或自行就醫的事故紀錄，因此實際因肇事影響死傷人數超過內政部警政署公佈數據。 肇事資料的不確定性、不完全性與模糊性，符合約略集合理論( rough sets theory )的概念，本研利用約略集合技術分析肇事特性，以民國94年內政部警政署登錄155,814件事故為範圍，推演肇事特性法則、肇事原因法則、一般與重大事故法則和保護裝備與行動電話法則並進行研析，獲得之法則經過驗證後顯示一般與重大事故法則整體判中率極高，顯示將約略集合技術應用於肇事資料分析，可以輕易進行包含重複紀錄、簡化屬性、發掘屬性關係並推演最適當的法則。 依據約略集合推演的肇事特性法則顯示，屬性類別影響大小為道路屬性、車輛屬性、環境屬性與當事人屬性。應用約略集合技術於肇事原因法則和一般與重大事故法則，結果顯示共同肇事屬性數量下，嚴重程度分類較多整體判中率較低；嚴重程度分類較少整體判中率較高。應用約略集合推演保護裝備與行動電話法則，結果顯示當事人受傷的可能性為96%、死亡的可能性為3%、造成部分人員死亡部分人員受傷的可能性為1%。 本研究以實際資料進行進行研究證實約略集合理論對於分析不確定性、不完全性與模糊性之肇事資料是個有用且有效率的工具，未來可針對法則擬定政策與改善措施，並建立肇事特性法則與改善對策查詢，協助我國交通事故主管機關將肇事管理系統資料庫中的資料作最佳化的運用。 Traffic accidents rose from 2001 to 2005. For example, 155,814 traffic accident in 2005 did not include they compromised or took medical treatment by themselves. Therefore, the number of casualties caused more than national police agency, ministry of the interior published data. Incomplete and ambiguous data of traffic accidents accord with rough set theory concept. The study use rough set theory to analyze the characteristics and rules of accidents to national police agency, ministry of the interior in 2005. Rough set theory can remove the redundant records, simplify the attribute-value table, discover the attribute association and induce the proper decision rules. Basing on the characteristics of the accident rules, attributes were influenced by attributes, vehicle attributes, environmental attributes and troublemaker attribute. The cause of accident and the rules of general and major accident had the same number of attributes, the classification is mickler the total accuracy is lower, the classification is less the total accuracy is higher. Applying rough set theory to the protective equipment and mobile phone rules show that troublemakers will be injured 96%, dead 3%, some injured and the others dead 1% probably. The study provides evidences showing that rough set theory is a useful tool for the analysis of incomplete and ambiguous data. The study was being conducted in which results from the whole set of data is presented and interpreted in order to obtain a better view of the condition of traffic accidents and be able to increase the effectiveness of the decision-making process.
