學校建築耐震能力資料探勘模型評估與系統建置

本文以資料探勘技術建置校舍建築耐震能力初步評判系統。探勘資料以教育部近年已完成的臺東縣中小學的272 棟耐震能力詳細評估資料新建為採礦資料庫，並以決策樹（decision trees）、貝式機率分類（bayes classifier）、及類神經網路（cellular neural network）等三種演算法共五種預測模型進行預測精準度評估，分析結果顯示決策樹演算法作為資料探勘模型，預測準確率可達93.82%，為五種模型中最佳的預測模型。再依最佳預測模型建置一套快速評估學校建築物耐震能力的系統，提供給學校建築管理人員在可獲得的建築耐震評估所需資料前提下，將未進行耐震能力評估的建築物資料輸入系統，讓學校能更快速的初步判斷建築物的耐震能力，進而減少震災發生時生命財產的損失。為再次驗證系統的可信度，本文採用其他學者之學校建築耐震評估資料輸入系統中測試，亦能得到相似的準確率，多重驗證本文成果具實用性。

關鍵字

典型校舍；資料探勘；預測模型

並列摘要

A preliminary School Building’s Seismic Capacity Evaluation System is established by using the data mining technique. The detailed seismic capacity evaluation reports of Taitung’s 272 elementary and high school buildings were used as the basis to establish a data mining data base. Three algorithms and five predictive models were used to evaluate the forecast precision of the seismic capacity of each building. The three algorithms are Decision Tree, Bayes Classifier, and Cellular Neural Network. The best model from the analysis results is the decision tree algorithm, whose accuracy rate has reached to 93.82%. A rapid school building’s seismic capacity evaluation system was then established based on the best fix model. The school building administrators can use this system to evaluate typical school building’s seismic capacity which has no detailed seismic capacity evaluation report. To check the applicability of the system, school buildings’ seismic detailed evaluation reports not from Taitung schools were also used and compared. The accuracy of this system was confirmed.