股票投資是人們生活中很常見的一種投資方式，而預測股價的漲、跌走勢，一直以來都是學者和金融研究人員努力研究的主題，如果能實現一種高效的股價預測方法，據此制定股票投資策略，就可以降低人們投資的風險，並增加投資收益。因此本研究利用類神經網路作為台灣加權股價指數之相關研究，嘗試用類神經網路去預測台灣加權股價指數未來的走勢，比較類神經網路是否能比傳統的Fama-French訂價模型得到更好的結果，再加入相關研究中常見的類神經網路模型，比較哪個結果誤差較低，其中包括類神經網路(Artificial Neural Network，以下簡稱ANN)，長短期記憶(Long Short Term Memory，以下簡稱LSTM)，兩個類神經網路模型。樣本期間為2011年到2020年，訓練期與預測期為5：1，使用python架構類神經網路模型，再加入其他因子於模型中，看是否能降低誤差，結果CNN-LSTM的誤差為最小。
Stock investment is a very common investment method in people's life, and predicting the rise and fall of stock prices has always been the subject of scholars and financial researchers. Investment strategies can reduce the risk of people's investment and increase investment returns. Therefore, this study uses the neural network as the relevant research of Taiwan's weighted stock price index, tries to use the neural network to predict the future trend of the Taiwan weighted stock price index, and compares whether the neural network can be compared with the traditional Fama-French Pricing model. For better results, add the common neural network model in related research, and compare which result has lower error, including artificial neural network (Artificial Neural Network, hereinafter referred to as ANN), Long Short Term Memory (Long Short Term Memory, Hereinafter referred to as LSTM), two types of neural network models. The sample period is from 2012 to 2021, and the training period and prediction period are 5:1. The python architecture neural network model is used, and other factors are added to the model to see if the error can be reduced. Initial results are the CNN-LSTM is the best.
