KTU Institute of Automation and Control Systems Intelligent

KTU Institute of Automation and Control Systems Intelligent systems in banking industry: survey and future Rimvydas Simutis Kaunas University of Technology, Lithuania Penkių Kontinentų Bankinės Technologijos (BS/2), Lithuania BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Outline q What are intelligent systems? q Why these systems are “hot” now? q Competitive edge and intelligent systems q Intelligent systems techniques q Where are we today? q Some application in banking sector q What is the future? q BS/2 and intelligent systems BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems q What are intelligent systems? BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Intelligent systems – a lot of discussions but none unified definition till now. Definition in our applications: Intelligent systems – systems which combine an a priori knowledge and real-time information (knowledge) for decision making Intelligence 1 A priori knowledge Extended knowledge Real – time knowledge Decisions 0 Knowledge used BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Three important components for design of intelligent systems Human Experts/Fundamentals – DATA – Evolution Fundamentals Improvements through Evolution Human experts A priori knowledge Extended knowledge Decisions Real – time knowledge DATA BS/2 Conference, Vilnius, 2008 June 13 Objectives

KTU Institute of Automation and Control Systems Why these systems are “hot” today? BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Computation power of today's computers is strong and is improving Basic techniques for design of intelligent systems are already developed Global communication and data basis systems are powerful Prices for human intelligence are increasing and are coming closer to “silicone intelligence” prices BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Computation power and point of “singularity” BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Competitive edge and intelligent systems BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Starting 2005 traditional IT Co are like utility companies (water supply, electricity supply - Internet supply building material warehouse - data warehouse. . etc) 2005 IT Companies Traditional IT Utilities (no competitive edge) IT based on Intelligent Systems Breakthrough in competition and innovation BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Task for innovative IT companies: switching part of recourses for developing of software and tools for automatic knowledge extraction and decision making Knowledge extraction tools Improvements through Evolution A priori knowledge Extended knowledge Decisions Objectives Real – time knowledge Knowledge extraction tools DATA BS/2 Conference, Vilnius, 2008 June 13 Decision making tools

KTU Institute of Automation and Control Systems Intelligent systems techniques BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from human experts q Formal models q Expert systems q Fuzzy expert systems, hierarchical fuzzy systems q Neuro – fuzzy systems q “Cloning” of human experts BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from human experts q Fuzzy systems BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from human experts q ‘Cloning’ of human experts - Experts are expensive; - Experts are always busy; - Experts are ‘critical recourses’; - Experts can leave the company; BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from human experts q ‘Cloning’ of human experts BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from data q Data mining techniques, advanced visualization q Decision tree techniques q Extraction of Fuzzy rules q Case based reasoning q Artificial neural networks q Self-organizing neural networks q Hierarchical learning of structures (Jeff Hawkins) --- Every day, week or month --BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from data q Decision tree techniques BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from data q Extraction of fuzzy rules BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from data q Artificial neural networks BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Extraction knowledge from data q Self-organizing neural networks BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Techniques for Decision Optimization q Traditional numerical optimization methods q Stochastic optimization q Evolutionary/genetic programming q Swarm intelligence, ant colony optimization q Multi-agent systems BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Techniques for Decision Optimization q Swarm intelligence, ant colony optimization BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Where are we today? BS/2 Conference, Vilnius, 2008 June 13

BS/2 Conference, Vilnius, 2008 June 13 Institute of Automation and Control Systems KTU

KTU Institute of Automation and Control Systems Some application in banking sector q In all applications only one or two components of the intelligent systems frame are used BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Application Prediction with ANN Classification with ANN Clustering with SOM Marketing and Sales Forecasting customer response (Bounds, 1997); Market development forecasting (Wang, 1999); Sales forecasting (Kong, 1995); Price elasticity modeling (Gruca, 1998) Target marketing (Zahavi, 1997); Customer satisfaction assessment ( Temponi, 1999); Customer loyalty and retention (Mozer, 2000, Smith, 2000) Market segmentation (Reutterer, 2000); Customer behavior analysis (Watkins, 1998); Brand analysis (Reutterer, 2000); Market basket analysis (Evans, 1997); Storage layout (Su, 1995) Credit scoring (West, 2001) Risk assessment (Garavaglia, 1996) Signature verification (Abu-Rezq, 1999) Risk Assessment and Financial health prediction (St. Accounting John, 2000); Credit scoring (Jensen, 1992); Insolvency prediction (Brockett, 1997); Compensation assessment (Borgulya, 1999); Bancrupcy classification (Wilson, 1997); Credit scoring (West, 2000, Long, 2000); Fraud detection (Holder, 1995, He, 1997); Signature verification (Ageenko, 1998) BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Application Prediction with ANN Classification with ANN Business Policy, Managements and Strategy Evaluating strategies (Chien, 1999); Assisting decision making (Wu, 1999) Impact of strategy on performance (St. John, 2000); Impact of management practices on performance (Bertels, 1999) Finance Hedging (Hutchinson, 1994); Futures forecasting (Grudnitski, 1993); FOREX forecasting (Leung, 2000); Investment management (Barr, 1994); Stock trend classification (Saad, 1998); Client authentication (Graham, 1988); Bond rating (Dutta, 1993) BS/2 Conference, Vilnius, 2008 June 13 Clustering with SOM Impact of strategy on performance (Biscontry, 2000); Assisting decision making (Lin, 2000) Economic rating (Kaski, 1996); Interest rate structure analysis (Cottrell, 1997) Mutual fund selection (Deboeck, 1998)

KTU Institute of Automation and Control Systems Fuzzy-enhanced score card system used in BMW bank BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems Portfolio management application using multi-agent system technology WARREN BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems What is the future? BS/2 Conference, Vilnius, 2008 June 13

BS/2 Conference, Vilnius, 2008 June 13 Institute of Automation and Control Systems KTU

KTU Institute of Automation and Control Systems q All depends on I_Price_Ratio = Silicone Intelligence Price/ Human Intelligence Price q Low I_Price_Ratio will stimulate development of advanced tools and software for intelligent systems and their applications q Combination of all three components of intelligent systems technique is crucial in the future (crucial for high IQ !) q Development of user friendly tools for design of intelligent systems is highest priority (basic methods are known) BS/2 Conference, Vilnius, 2008 June 13

KTU Institute of Automation and Control Systems BS/2 and intelligent systems q Application of multi-agent systems and optimization techniques for decision making q Application of artificial neural networks, support vector machines and associative neural networks for knowledge extraction from data q Combination of expert knowledge/data knowledge for warehouse operations BS/2 Conference, Vilnius, 2008 June 13