Visual knowledge discovery and machine learning for investment strategy

Knowledge discovery is an important aspect of human cognition. The advantage of the visual approach is in opportunity to substitute some complex cognitive tasks by easier perceptual tasks. However for cognitive tasks such as financial investment decision making this opportunity faces the challenge that financial data are abstract multidimensional and multivariate, i.e., outside of traditional visual perception in 2D or 3D world. This paper presents an approach to find an investment strategy based on pattern discovery in multidimensional space of specifically prepared time series. Visualization based on the lossless Collocated Paired Coordinates (CPC) plays an important role in this approach for building the criteria in the multidimensional space for finding an efficient investment strategy. Criteria generated with the CPC approach allow reducing/compressing space using simple directed graphs with beginnings and the ends located in different time points. The dedicated subspaces constructed for time series include characteristics such as Bollinger Band, difference between moving averages, changes in volume etc. Extensive simulation studies have been performed in learning/testing context. Effective relations were found for one-hour EURUSD pair for recent and historical data. Also the method has been explored for one-day EURUSD time series n 2D and 3D visualization spaces. The main positive result is finding the effective split of a normalized 3D space on 4 × 4 × 4 cubes in the visualization space that leads to a profitable investment decision (long, short position or nothing). The strategy is ready for implementation in algotrading mode.