基於過去的預測模式對股利宣告效應的忽視，本研究主要目的在於利用股利宣告 事件中的異常報酬（ Abnormal Return ），作為股利宣告事件的影響值（效應）， 用以改 進目前預測模式之準確性。本研究利用現存的兩個股價預測模式作為基本模式，在基本模式 中加入股利宣告事件影響值，以改進現存股價預測模式之不足。本研究之股價預測模式採用 類神經網路（ Artificial Neural Network ）為主要架構， 隨機抽樣 30 家臺灣上市公司 為研究樣本，研究期間為民國七十年至民國八十五年，經由統計結果顯示本研究所提之股價 預測模式確能顯著改進誤差，同時對漲跌預測也獲得明顯改進。

　　　　　In the past, the effect of the dividends announcement event is always ignored and not addressed in existing stock price forecasting models. In this study, we employ the abnormal returns of the dividends announcement event as effect values for forecasting stock price on the Taiwan stock market. This study proposes new stock price forecasting models combining existing forecasting models and the information of abnormal returns. These models are developed and implemented using artificial neural networks (ANNs). The results based on a sample of 30 firms listed on the Taiwan Stock Exchange (TSE) for the period 1981-1996 are investigated. Conclusions obtained from this study are as follows: the proposed models with the information of abnormal returns are better than those without the information of abnormal returns in terms of absolute percentage error (APE) and stock price trend prediction.