應 Data Mining建置一分類模型

Applying Data Mining to Build a Classify Model

本研究以資料挖掘（Data Mining）、資料倉儲（Data Warehouse）以及分類（Classification）演算法規則中之決策樹（Decision-Tree）為基礎之分類法做為本研究之工具，以心臟科麻醉部為例，從醫療單位對抗生素進行導管菌落判斷感染監控的歷史案例中，建構出可用以預測心靜脈導管塗抹抗生素在病患身上的作用結果之分類模式，用以協助醫藥人員有效判斷感染情形與特徵，進而提升抗生素之用藥醫療品質、降低可能的醫療資源浪費。實驗結果顯示，本研究所收集的導管菌落判斷感染監控資料中有顯著類別間資料數量不對稱的現象，而造成模式預測偏向的問題。經由成本敏感法的運用預測錯誤成本的改善調整，提升分類預測的效能，建構出對各類別資料都有良好預測效能的理想模式。目的做為日後與專家系統之結合，發展出一套強而有力之醫療專家系統，減少醫療費用的浪費，同時在照顧病人品質方面能夠維持一定的水準。

The research use Data Mining, Data Warehouse and Decision-Tree that Classification perform the algorithm makes for the tool. Take department that anaesthetizes division of cardiology as an example. Build and construct out measurable Central Venous Catheter to apply in the historical case of judging Central Venous Catheter Insertion from the antibiotic and infecting Wipe the categorized model on patient of antibiotic. The paper use under-sampling, over-sampling and Part to solve skewed class distribution problem. And then use C4.5 algorithms to build up classifier. Verify via the expert that use this classifier can effectively help the medical expert to extract out it about Central Venous Catheter Insertion good knowledge rule.