Discovering Financial Investment Strategy through Wavelet-based SOM Networks

Recently, the recession of the global economy drives the coming of a new era of low interest rates, which resulted in the stock market as an alternative investment channel for investors. The diversity and complication of domain knowledge existing in the stock market make it very difficult for investors to make an effective investment decision. In the literature, it has been shown that intelligent decision support models can alleviate the burden of investors. In this study, we are in an attempt to develop such model by applying the over-whelming selforganization map (SOM) network. In contrast to most similar approaches toward the problem of stock prediction, we investigate the feature extraction issue in dealing with stock time series using discrete wavelet transform. We design a two-level SOM network and perform a novel trajectory analysis so that both short-term and long-term prediction can be provided. In addition, we not only endeavor ourselves to improving the accuracy of uncovering trading signals, but also to maximize the profits of trading. The experimental results demonstrate that the proposed decision model can help investment make the most profitable decisions.