Big Data Mining

The Big Data & Big Data Mining research focuses on developing scalable infrastructure and algorithms for discovering new and interesting knowledge hidden in large data bases. For this purpose, we have developed cluster-parallel software that utilizes the power of multi-core and many-core clusters of computers for the efficient distribution of in-memory compute-intensive tasks related to knowledge discovery from data. On top of this infrastructure, we have developed several novel big data mining algorithms that optimally leverage the developed infrastructure, that find applications in a large and diverse set of sub-domains, ranging from recommender systems engines to estimating individual consumers’ reservation price for specific products etc.

Our current infrastructure consists of a scalable parallel/distributed optimization library in Java (https://github.com/ioannischristou/popt4jlib) which provides both shared-memory as well as distributed/cluster-parallel computation APIs for developing efficient large-scale data mining and optimization algorithms. On top of that, several highly scalable algorithms and systems have been implemented, including

  • AMORE: parallel hybrid recommender systems engines
  • QARMA: parallel quantitative association rule mining algorithms
  • EXAMCE: data clustering of domains requiring many partitions
  • classifier ensembles combining decision trees and support vector machines,
  • and a large base of meta-heuristic optimization algorithms for large-scale optimization involving thousands of variables.

The applications of the developed technology are very diverse, and have been of use to many existing companies.

Eyes-On: Commercial fraud detection technology real-time application. Based on data clustering techniques for detecting outliers in data bases, a novel system has been developed that detects possible fraud-case scenarios in the production data of a world class player in the Lotteries and Games of Chance industry.

AMORE: Personalized movie recommendations for subscribers to Triple-Play services. Developed a production system that produces several times a day personalized recommendations for the subscribers to the “My Video Club” service of a major Triple-Play service provider in Greece.

i-Pricer: Based on scalable quantitative association rule mining technology, we have developed a commercial-strength system that can derive accurate upper bounds on the “reservation price” of individual repeat customers of a retailer for specific products. The product became the foundation for the formation of the Big Data start-up company “intelprize”.

We are also actively researching data mining of big-data time-series arising from many applications including financial time-series such as stock prices or forex markets. The developed systems are being tested in real money accounts with very promising results.

Patents

Dimitriadis, I.T. Christou, M. Bakopoulos: METHODS AND A SYSTEM FOR DETECTING FRAUD IN BETTING AND LOTTERY GAMES. Patent Granted Mar. 19, 2013. Patent#: US8401679 B2 (www.google.ca/patents/US8401679)

Selected Publications

I.T. Christou, E. Amolochitis, Z.-H. Tan, “QARMA: A Parallel Algorithm for Mining All Quantitative Association Rules and Some of its Applications”, Knowledge & Information Systems, Accepted with Minor Revisions, 2017

Goumatianos, I.T. Christou, P. Lindgren, R. Prasad “An Algorithmic Framework for Frequent Intraday Pattern Recognition and Exploitation in Forex Market”, Knowledge and Information Systems, DOI 10.1007/s10115-017-1052-2, 2017

I.T. Christou, E. Amolochitis, Z.-H. Tan, “AMORE: Design & Implementation of a Commercial-Strength Parallel Hybrid Movie Recommendation Engine”, Knowledge and Information Systems, 47(3), pp. 671-696, 2016

Goumatianos, Ioannis T. Christou, P. Lindgren, “Integrating Grid Template Patterns and Multiple Committees of Neural Networks in Forex Market”, Proc. 4th Computer Science Online Conf., Tomas Bata University, Zlin, Czech Republic, Apr. 27-30, 2015. In: Springer Series on Advances in Intelligent Systems & Computing

Amolochitis, Ioannis T. Christou, Z.-H. Tan, “Implementing a Commercial-Strength Parallel Hybrid Movie Recommendation Engine”, IEEE Intelligent Systems, 29(2), pp. 92-96, 2014.

Amolochitis, Ioannis T. Christou, Z.-H. Tan, R. Prasad, “A Hierarchical Heuristic for Efficient Re-ranking of Academic Search Results”, Information Processing & Management, 49(6), pp. 1326-1343, Nov. 2013

Goumatianos, Ioannis T. Christou, P. Lindgren, “Stock Selection System: Building Long/Short Portfolios Using Intraday Patterns”, Proc. Intl. Conf. on Applied Economics (ICOAE 2013), Istanbul, Turkey, June 2013. Published in Elsevier Procedia in Economics and Finance, vol. 5, pp. 298-307

Goumatianos, Ioannis T. Christou, P. Lindgren, “Useful Pattern Mining on Time Series: Applications in the Stock Market”, 2nd Intl Conf. on Pattern Recognition Applications and Methods (ICPRAM 2013), Barcelona, Spain, Feb. 15-18, 2013

Ioannis T. Christou, G. Gkekas, A. Kyrikou, “A Classifier Ensemble Approach to the TV Viewer Profile Adaptation Problem”, Intl Journal of Machine Learning and Cybernetics, 3(4), pp. 313-326, 2012

Ioannis T. Christou, M. Bakopoulos, T. Dimitriou et al, “Detecting Fraud in Online Games of Chance & Lotteries”, Expert Systems with Applications, 38(10), pp. 13158-13169, 2011.

Ioannis T. Christou, “Coordination of Cluster Ensembles via Exact Methods”, IEEE Transactions on Pattern Analysis and Machine Intelligence, 33(2), pp. 279-293, 2011

Menkovski, Ioannis T. Christou, and S. Efremidis, “Oblique Decision Trees using Embedded Support Vector Machines in Classifier Ensembles”, IEEE Cybernetic Intelligent Systems Conf., London, UK, Sep. 2008

Head of the Big Data Mining group