本研究主要以深度學習之遞迴類神經網路方法，實作具時間序列性之LSTM(Long Short-Term Memory)長短期記憶遞迴類神經網路（Recurrent Neural Network，RNN）模型，接收來自產業工廠即時製造過程中排放工業廢水監測數據，經本研究設計的實驗預測未來股價趨勢，並依據研究標的公司於台灣證券交易所公開之股價，驗證各期資料訓練出模型之預測準確率。 並提出產業製程排放廢水量與下期營收成正相關之假說，以科學實驗方法驗證假說可靠度。 本研究貢獻，主要驗證以當期廢污水排放各項監測數值及對應之股價，加以訓練類神經網路模型，進而以當期數據去預測未來股價趨勢，以評定本研究所創建的類神經網路模型之準確率及假說。 最終期望創建出具備領先性的“生產資源消耗面”之非財務性科技預測指標—新河指標，提供投資者作為研析標的公司未來股價走勢之依據。;This research uses Deep Learning technology, LSTM Network, to solve the prediction issue of future stock price. In contrast to traditional methods, it uses industrial wastewater dataset to train LSTM model. In experiment, it is designed to different models by deferred periods of the affected stock price and finds the most accurate model for stock price prediction. Moreover, this paper designs experiments to ascertain the hypothesis, industrial wastewater of factories influencing its future stock price trend, whether they have the positive correlation. The contribution of this research proves the future stock price prediction of manufacturing industry can use the leading index, industrial wastewater, effectively. And it also finds out using industrial wastewater dataset to intensify the accuracy of LSTM network in stock price prediction is a useful way. Ultimately to produce a non-finance leading index of stock prediction, New River index, by LSTM approach that helps investors to judge investment in advance is this research contribution.
