Parallel SAX/GA for financial pattern matching using NVIDIA's GPU

Abstract This paper starts by presenting a study from a computational performance standpoint of SAX/GA, an algorithm that uses the Symbolic Aggregate approXimation (SAX), to dimensionally reduce time series, and the Genetic Algorithm (GA) to optimise market trading strategies. This study highlights how the sequential implementation of SAX/GA and genetic operators works. This study is later used as the baseline for the development of parallel techniques capable of exploring the identified points of parallelism that simply focus on accelerating the heavy-duty fitness function to a full Graphical Processing Unit (GPU) accelerated GA. The implemented solutions accelerated the sequential single-core SAX/GA solution in about 30 times with a maximum of nearly 180 times.

[1]  Dan Roth,et al.  Efficient Pattern-Based Time Series Classification on GPU , 2012, 2012 IEEE 12th International Conference on Data Mining.

[2]  Byoung-Tak Zhang,et al.  Comparison of Selection Methods for Evolutionary Optimization , 2000 .

[3]  KeoghEamonn,et al.  Time series shapelets , 2011 .

[4]  Amer Bakhach,et al.  Forecasting directional changes in the FX markets , 2016, 2016 IEEE Symposium Series on Computational Intelligence (SSCI).

[5]  Tak-Chung Fu,et al.  Stock time series pattern matching: Template-based vs. rule-based approaches , 2007, Eng. Appl. Artif. Intell..

[6]  Achilleas Zapranis,et al.  Identification of the Head-and-Shoulders Technical Analysis Pattern with Neural Networks , 2010, ICANN.

[7]  Kyoji Kawagoe,et al.  Extended SAX: Extension of Symbolic Aggregate Approximation for Financial Time Series Data Representation , 2006 .

[8]  Wen-Shiung Chen,et al.  High performance data compression method with pattern matching for biomedical ECG and arterial pulse waveforms , 2004, Comput. Methods Programs Biomed..

[9]  Tzung-Pei Hong,et al.  Time series pattern discovery by a PIP-based evolutionary approach , 2013, Soft Comput..

[10]  Edward P. K. Tsang,et al.  Backlash Agent: A trading strategy based on Directional Change , 2016, 2016 IEEE Symposium Series on Computational Intelligence (SSCI).

[11]  Chonghui Guo,et al.  Similarity measure based on piecewise linear approximation and derivative dynamic time warping for time series mining , 2011, Expert Syst. Appl..

[12]  Nuno Horta,et al.  Multi-dimensional pattern discovery in financial time series using sax-ga with extended robustness , 2013, GECCO '13 Companion.

[13]  Wolfgang Kastner,et al.  Analysis of Similarity Measures in Times Series Clustering for the Discovery of Building Energy Patterns , 2013 .

[14]  Nuno Horta,et al.  A SAX-GA approach to evolve investment strategies on financial markets based on pattern discovery techniques , 2013, Expert Syst. Appl..