Mining in-depth patterns in stock market

Stock trading plays an important role for supporting profitable stock investment. In particular, more and more data mining-based technical trading rules have been developed and used in stock trading systems to assist investors with their smart trading decisions. However, many mined trading rules are of no interest to traders and brokers because they are discovered based on statistical significance without checking traders' interestingness concerns. To this end, this paper proposes in-depth data mining technologies to overcome the disadvantages of current data mining methods. We implement a decision support in-depth trading pattern discovery system with Robust Genetic Algorithms (RGA). The system integrates expert knowledge and considers domain constraints into the trading rule development. We further utilise this technique to mine actionable stock-rule pairs targeting behaviour with high return at low risk. The proposed approaches are tested in real stock orderbook data with varying investment strategies.