本研究提出一個新穎的整合模型，此模型整合支援向量機與整體學習機制應用於預測公司績效。整體學習機制之採用可有效降低預測偏誤，以及提升模型預測品質。本研究資料取自台灣經濟新報資料庫(TEJ)，研究對象係以2010年到2012年之台灣上市上櫃電子業為主。實證結果顯示非財務變數(公司治理變數)對於公司績效預測具有重要性，且支援向量機整合整體學習機制之模型具有相當優越預測能力。本研究結果將可進一步提供予管理者進行公司績效評估，有利於在公司有限資源情況下，調整公司策略，進行公司績效改善。
In this study, we proposed a hybrid mechanism that incorporated support vector machine (SVM) and ensemble learning to forcast the firms’ operating performance. The SVM has been demonstrated it outstanding generalization ability. The ensemble learning is used to increase the forecasting performance and decrease the forecasting variance. Grounded on ensemble learning, the forecasting quality of hybrid mechanism can be enhanced successfully. The research data was gathered from Taiwan Economic Journal (TEJ) and the research catagoried was focused on publicly electronic industries. The research period was ranged from 2010 to 2012. According to our rearch finding, the non-financial indicatore, that is corporate governance indicators, play essential roles in forecasting quality improvement. The SVM based on ensemble strategy can be complemented error made by singular component. The research finding is examined by real cases and the result can be viewed as guideline for managers to modify the corporate operating direction and allocate the limited economic resources.
