This project predicts stock performance, either stock price or trend, based on three machine learning algorithms: Support Vector Regression (SVR), Long short-term memory (LSTM) and Random Forest (RF). We construct our own dataset by combining calculated 12 technical factors and fundamental factors of 7 stocks, and pre-process the data before applying them to specific models by normalizing and dimension reduction techniques. We test the performance of SVR and LSTM by root mean squared error (RMSE) while RF by accuracy. After the experiments, SVR model performs better than LSTM in predicting stock price. RF model shows relatively high accuracy in predicting the price trend and the performance closely relates to choice of class and decision path.
Chengshuo Zhang cszhang@umich.edu Shouren Wang shourenw@umich.edu
Yanlin Xiao yanlinx@umich.edu Minghao Yang minghaoy@umich.edu