查詢結果分析
相關文獻
- 我國財產保險公司償付能力預警模型研究
- 群眾募資平臺動態預測之研究:以Kickstarter平臺為例
- 在國際財務報導準則IFRSs下以Crisp-DM流程建立財務風險管理平臺
- Churn Prediction Based on the Analysis of Customers' Preferences and Social Behavior on a Big Data Platform
- 醫療可及性是否左右房價?機器學習之迴歸樹及隨機森林模型的應用
- 應用資料探勘技術於信用卡逾期繳款模型之研究
- 應用Python的Sklearn機器學習預測臺北市老舊聚落防災策略
- 以二元羅吉斯迴歸與機器學習模型定義之自駕車車禍嚴重程度相關因子比較
- 資料挖掘的運用--以決策樹方法分析保送甄試學生入學四技二專報到情形
- 應用決策樹的資訊管理--護理人員的排班
頁籤選單縮合
題名 | 我國財產保險公司償付能力預警模型研究=The Study of Insolvency Prediction Models for Chinese Property Insurance Companies |
---|---|
作者 | 熊逸柳; 伍向豪; 熊逸柳; 伍向豪; Xiong, Yi-liu; Wu, Shiang-hau; Xiong, Yiliu; Wu, Shianghau; |
期刊 | 澳門科技大學學報 |
出版日期 | 20150600 |
卷期 | 9:1 2015.06[民104.06] |
頁次 | 頁74-87 |
分類號 | 563.727 |
語文 | chi |
關鍵詞 | 財產保險公司; 償付能力預警; 決策樹; 隨機森林; 固定效應模型; Property insurance company; Insolvency prediction; Decision trees; Random forests; Fixed effect model; |
中文摘要 | 保險公司作為我國金融業的“四大支柱"之一,其經營穩定性對國民經濟穩定發展具有重要影響。近年來,隨著保險行業的不斷創新與發展,保險公司所面臨的風險也更加複雜多樣,償付能力不足狀況時有發生。本文搜集了2009~2012年24家中資財產保險公司的財務數據,結合國內外償付能力預警研究及我國償付能力監管指標,選取了10個影響償付能力的因素作為預警指標。首先將財產保險公司按照償付能力充足率是否達到100%分為兩類通過面板數據Logistic回歸模型進行分析,發現再保險率對償付能力充足率有顯著的正向影響。之後,將財產保險公司按照償付能力充足率是否達到100%分為兩類,運用R軟件中的Rattle包進行決策樹和隨機森林的預警模型構建,都顯示出了很好的預測效果,但隨機森林模型的泛化能力更強。繼而,依據現實情況將保險公司分三類監管的要求,本文又構建了三分類隨機森林模型,對測試數據的預測準確率達93.75%。 |
英文摘要 | As one of the “four pillars” in Chinese financial industry, insurance companies have the essential influence on the stable development of national economy. With the growth of insurance industry, its external risk has become a more challenging problem, and the insolvency issue occurs. In this paper, financial data of 24 Chinese property insurance companies were collected from the year 2009 to 2012. In consideration of the past study of insolvency prediction and the solvency regulatory indices, 10 factors that had impact on insolvency were selected as relative variables. Then the result of Panel Data Logistic Regression Model showed reinsurance raio had significant positive effect on solvency. In order to make the classification analysis, property insurance companies were divided into two categories depending on whether its solvency ratio reached 100%. The Rattle package in R was applied to build Decision Trees and Random Forests Model, which all showed satisfied prediction results, but the generalization ability of Random Forests was better. Then, referring to the actual regulatory requirement, property insurance companies were grouped into three classes to construct a Random Forest Model, the accuracy of prediction of this model is 93.75%. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。