金融風暴後,台灣許多體質不健全的公司陸續爆發財務危機引起股市重挫,造成了投資者的重大損失。早期的股市交易評價理論及傳統的企業體質檢定模型的評價模式皆過於簡化,利用簡單線性關係來推導風險與報酬的關係,不但架構過於簡單且使用上有許多限制條件。然而實際的金融市場是動態且具有高度複雜性,使其使用上可能無法滿足這些限制,因此本論文以自組織映射圖神經網路建構出非監督式學習與視覺化功能的模型以分析各公司體質變化之情形。預測出未來最有可能轉變的方向,期望不受限於傳統模型的諸多條件,利用自組織映射圖神經網路具有非監督式分群的學習能力及視覺化聚類的優點,以動態的構面對企業財務體質進行分群檢定,將企業體質漸變過程投射於二維平面,並分析其落點位置與體質移動軌跡的趨勢,區分體質優良與不良的企業以進行投資。最後實驗結果指出,自組織映射圖網路,可以有效區分出體質優良的企業與體質不良的企業,並以其公司落點位置與移動軌跡,判別該公司是否屬於成長趨勢,或者屬於衰退趨勢,以提供投資者進行投資決策。
After Asia Crisis, many unhealthy companies busted out emerging financial crisis and led to the crash of the stock market. The evaluation method of early stock trading evaluation theory and enterprise constitution examination model are too simplified, the structure of using single linear relationship to derive relationship between risk and reward is too simplified or have several constraints. However the real financial market is dynamic and highly complicated, the constraints on evaluation cannot be satisfied, hence this study uses Self-Organizing Map (SOM) to build an unsupervised learning and visualized model to analysis the status of different enterprise physical constitution changes. Forecasting the most probable direction of change without the constraints of traditional models and using the advantage of unsupervised and visualized clustering of SOM, enterprise’s financial constitution can be clustered dynamically for examination, the gradual change of enterprise physical constitution is projected on two-dimensional surface and its placement position is analyzed along with the trend of moving orbit of physical constitution Hence, enterprise physical constitution is differentiated into good or not good for investment decision accordingly. Final experiment indicated that SOM effectively ditinguished whether the physical constitution of enterprise is excellent or awful, and the enterprise is differentiated into growing or declining trend based on its placement position and moving orbit for investors to make investment decision.