過去關於預測相關之研究大多將焦點集中在財務危機與信用風險等主題，對於企業營運表現預測模型卻鮮少著墨，但企業營運表現不彰是導致企業財務危機的主要因素。有關企業營運表現之主要評測指標大多以資產報酬率與股東權益報酬率為替代變數，但此評測指標是屬於單一投入與單一產出變數之機制，採用此簡易的評測模型來評斷企業營運表現將可能產生可信度之疑慮，為了減緩此問題，可以同時處理多個投入變數與多個產出變數之資料包絡分析法被提出，但此傳統模型在提供學習參考集合給不具效率之企業進行學習時，往往提供企業無法達成之目標，為了可以有效克服傳統資料包絡分析法之問題，本模型將採用情境相依資料包絡分析法，此模型針對不具備效率之企業提供一個其可以達成之目標與學習方向。此外，企業營運績效過去的評測模型之演進與強化，仍缺乏考慮文字型態資料，為了建構更加完善之評測模型，本研究採用企業社會責任報告書之文字資訊，並彙整財務報表之數值資訊，進而將其結果匯入人工智慧技術，最後建構出企業績效之預測模型。經實證結果顯示，本研究模型是一個相當可靠的預測模型，可以提供管理者對於未來營運決策與投資組合調整之輔助機制。
Compared with well-defined researches, such as financial crisis prediction or credit rating forecasting, the works on performance evaluation that has been widely deemed as a main trigger for financial difficulties, is not received sufficient focuses by researchers. Returns on assets (ROA) and returns on equity (ROE) with the merits of intuitiveness and easy-to-use which turn out to be the most adopted measures in per-formance evaluation. However, the mentioned above measures are classified into one input and one output category that are unable to depict the real situation surrounded by the companies, especially in Today’s highly turmoil atmosphere. To confront this, data envelopment analysis (DEA) that can handle multiple input and multiple output varia-bles simultaneously without pre-determining a cost function is considered. The prob-lem of conventional DEA is that the reference benchmarks for inefficient decision making units (DMUs) sometimes ends up with targets that are complicated to attain. To combat this, the context-dependent DEA (CD-DEA) is considered. That is, the ref-erence benchmarks yielded by CD-DEA for inefficient DMUs is reachable. Even the advanced performance evaluation models have been proposed, these models still can-not explain the current and future performance very well. One of the possible reason is lacking of considering narrative messages which can be explored considerably from textual-based documents (i.e., corporate social responsibility (CSR) reports). To deal with this obstacle, this study extracts the readability form CSR’s reports and collect numerical information from financial statements and then fed the analyzed outcome into an artificial intelligence (AI) technique to construct the performance forecasting model. By doing so, we can examine the impact of narrative messages on forecasting model’s performance. The managers can view this framework as a decision support system to formulate their future policies and adjust their investment portfolios.
