Grant: Idea to Innovation

As an investor, how can you protect yourself against the uncertainty of today's volatile market? To answer this question, investors, large and small, fund managers and banks, increasingly rely on complex algorithms and high-performance computer systems in order to analyse large amounts of real-time financial data. The financial market-miner system (MarketMiner) is a component-based workflow environment using correlation and clustering of stock market data to construct a dynamic model that captures relationships among all stocks in the market as it evolves through time. MarketMiner leverages state-of-the-art parallel algorithms for covariance/correlation and clustering as the basis of a highly scalable and robust financial analytics engine, enabling novel capabilities for application areas such as risk management, intelligent portfolio selection and proprietary sentiment indicators. The system can be used offline to identify past trends or used online for streaming high-frequency intra-day trading data. As proof of concept, three applications (RiskView, Sentiment View, and PortfolioFinder) were developed based on the analytics from MarketMiner. We intend to experimentally demonstrate in financial terms the value of these applications and thus the competitive advantage of using MarketMiner analytics. In today's market, to out-compete the competition one will need to out-compute them. The trend towards automated trading and the availability of online data at sub-millisecond rates introduces new demands and opportunties which require quick online analysis and decision processing. The need to perform more computation faster comes at a time when there is a paradigm shift in the computing industry to parallelism with the use of computing clusters and multicore architectures. Marketminer is at the nexus of these two technological trends and presents an opportunity to develop expertise in these areas and help keep Canadian financial industy companies competitive.