Abstract:
In the past few years, the neural network models have been widely applied in many fields such as finance, medicine, geology and physics. In the financial industry, the predictive ability of neural network models has proven highly effective and accurate in recent years. However, forecasting accurately the stock indexes in reality is more difficult. In different stock markets and during great volatility in the stock market, different models produce different performance. In this thesis, I use 3 widely used neural network models: Multilayer Perceptron (MLP), Long Short Term memory (LSTM) and Convolutional Neural Network (CNN) to forecast stock indexes based on historical data. The input variables used in the model include three main groups: daily trading variables, technical variables and macroeconomic variables. I use two data sets including the SP500 index, the VNINDEX index from two different stock markets. In this thesis, I will compare the accuracy of each model and indicate which data sets are more predictable.
