Data-Based Stocks Selection Method via Revised Threshold Accepting Algorithm

In portfolio management, stocks selection and their weights decision are the key issues. A typical fund manager would use fundamental or technical analysis to select the stocks at first, and then apply certain models, e.g. Markowitz model, to determine the proper weight for each stock. However, this subjective method in selecting stocks is heavily affected by the fund managers´ knowledge, ability and luck. In this paper, we propose an objective data-based approach in selecting the stocks in portfolios. We fully study the data of all stocks in S&P 500 and find outstanding portfolios with 25 or less stocks. We develop a revised threshold accepting global optimization algorithm which can efficiently deal with the large computation here. We test our portfolios with the latest stock price changes and show that they performs much better than the S&P 500 index.