dc.description.abstract | The goal of portfolio management is to allocate the limited money into multiple securities effectively to earn more money. The two key factors of obtaining high profit are “the trading time” and “asset allocation”. Most of the past researches focus on one specific problem, for example, trading signal prediction, asset allocation, adaptive investment goal setting etc. But in real investment, investors need solutions for all these problems to achieve investment goal. This research combines trading signal prediction [9], Time Invariant Portfolio Protection [10], Optimal Dynamic Asset Allocation [20] into a portfolio management system. In the first part, we use turning points to partition the stock price, and use Back-propagation neural network (BPNN) to learn and predict the trading signal for each stock. We propose a new way to calculate the trading signal to avoid parameter tuning required in [9]. The second part is asset allocation. With asset protection mechanism (TIPP), we set an explicit ROI as the investment goal. We then allocate money for investment target for their probability to reach the investment goal in each rebalance. The experiment shows that the new way to calculate trading signal has similar performance with the original method but avoids the parameter tuning problem. Furthermore, with asset protection mechanism, our portfolio management system would receive less damage in encountering bear market. However, the fixed investment return goal would limit the profit in bull market. Therefore, we start a new round when the portfolio reaches the investment goal, and successfully makes the portfolio management system conquer the problem in bull market. For long-term investment, this mechanism could get better performance by setting appropriate return rate. | en_US |