ERM Symposium 2012 Washington, D.C.

Transcription

1 ERM Symposium 2012 Washington, D.C. Jefferson Braswell Tahoe Blue Ltd 4/19/12 1

2 4/19/12 2

3 4/19/12 3

4 4/19/12 4

5 4/19/12 5

6 4/19/12 6

7 Requires the extraction of information and associations from data in order to attempt to predict future trends and behavior patterns. Involves the identification of relationships between explanatory variables and the predicted variables based on past occurrences in order to anticipate or predict future outcomes. 4/19/12 7

8 Predictive models Credit scoring, fraud detection Customer relationship management (CRM) Risk Management, Underwriting Regressions, Support Vector Machines, Supervised Learning Descriptive Models Categorization and Classification, Cluster Analysis Development of Simulation Sub-Models Data Mining, Unsupervised Learning Decision Support Optimization, Maximization/Minimization, Objective Functions Many variables and factors, including outputs of other models 4/19/12 8

9 Data Mining Anomaly Detection Association Rule Learning Clustering Classification Regression Summarization Regression Models Ordinary Least Squares Linear Regression models Discrete Choice models Time Series models Survival or Duration models Supervised Learning Naïve Bayes Neural networks Radial basis functions Support Vector Machines Linear SVM Multiclass SVM Structured SVM Gradient boosting K-nearest neighbors Random Forest Random Multinomial Logit Random Naïve Bayes Geospatial predictive models Group Method of Data Handling 4/19/12 9

10 4/19/12 10

11 4/19/12 11

12 4/19/12 12

13 Source: Banking and the Real Economy, by Dr. Robert Mark and Jefferson Braswell, in the forthcoming Handbook of Financial Risk Information, to be published by Cambridge University Press 4/19/12 13

14 Source: Banking and the Real Economy, by Dr. Robert Mark and Jefferson Braswell, in the forthcoming Handbook of Financial Risk Information, to be published by Cambridge University Press 4/19/12 14

15 Distribution of Operational Losses by Frequency and Magnitude Source: CRUZ, M.G Modelling, measuring and hedging operational risk, John Wiley & Sons Ltd 4/19/12 15

16 Supervised Learning Business Unit Op Risk Score 4/19/12 16

17 Thought recognition is the application of pattern recognition and machine learning methods to discriminate patterns of brain activity associated with a particular cognitive state or physiological condition. - Marc M. Palatucci Source: Pages 2-4, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 17

18 Source: Page 27, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 18

19 Source: Page 36, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 19

20 Page 17, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 20

21 Source: Page 26, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 21

22 Source: Page 75, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 22

23 Source: Page 59, Thought Recognition: Predicting and Decoding Brain Activity Using the Zero-Shot Learning Model, Doctoral Thesis by Marc M. Palatucci, April 25, 2011, Robotics Institute School of Computer Science, Carnegie Mellon University 4/19/12 23

24 4/19/12 24

25 Jefferson Braswell 4/19/12 25 Washington, D.C.