The Application of Decision Tree to Analyze the Financial Health of Non-Life Insurance Industry in Taiwan

[[abstract]]本研究採用決策樹演算法探討2005年至2009年台灣16家產險公司財務健全度，以2005年至2007年的財務資料作為訓練資料，2008至2009年財務資料作為測試資料，建立財務健全度預警模型，並找出重要影響變數。首先採用全部20個財務比率變數應用決策樹分析建立預警模型，然後利用因素分析，將二十個財務比率整理成四個財務指標，分別為「整體營運績效指標」、「獲利能力指標」、「清償能力指標」及「投資績效指標」，並以因素得點做決策樹分析。 研究結果發現，採用全部變數所建立的財務健全度預警模型有較好的正確預測率，整體正確預測率在85%以上，而財務健全度預警模型的主要影響變數為資產報酬率、負債對流動資產比率、資金運用淨收益率、自留綜合率。[[abstract]]The purpose of the research aimed to explore financial health from 16 non-life Insurance companies from 2005 to 2009 by using Decision Tree Algorithm taking financial variables from 2005 to 2007 as training data set and financial variables from 2008 to 2009 as test data set Both training data set and test data set constructed a financial health prediction model to discover the critical variables which influenced financial health First the study initially adopted twenty financial ratio variables and applied decision tree analysis to construct a prediction model secondly factor analysis are used to categorize twenty financial ratios into four financial indices: “whole operation and management efficiency index ” “benefit ability index ” “insolvency index” and “invest performance index ” Then factor scores were led to decision tree analysis The result of the study revealed that financial health prediction model which applied whole variables with better accuracy ratio of predicting The whole correct predicting rate is above 85% The critical variables of financial health prediction model are “return on assets ratio ” “liabilities to liquidity assets ratio ” “earnings ratio of fund application” and “combined ratio