In this project, we conduct fundamental analysis and prediction of stock returns with respect to different sections with neural network technology using 16 factors as input predictors. Quarterly stock returns in 2017 are used as the output vector. Portfolio is then generated according to forecast result by neural networks and rebalance at the end of each quarter. It yields an annual abnormal return over 65% on average. Significant abnormal returns indicate the value of neural network as an efficient tool for fundamental analysis and forecasting in the US markets.
Use ANN to predict future returns of stocks based on fundamental data.
Based on predicted returns, construct a number of investment portfolios.
Evaluate the portfolios by comparing with the S&P 500 index, showing the portfolios outperform the index.