This Title All WIREs
How to cite this WIREs title:
WIREs Data Mining Knowl Discov
Impact Factor: 7.250

Inference enterprise models: An approach to organizational performance improvement

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

We demonstrate that our success in solving a set of increasingly complex challenge problems is associated with an inference enterprise (IE) using inference enterprise models (IEMs). As part of a sponsored research competition, we created a multimodeling inference enterprise modeling (MIEM) process to achieve winning scores on a spectrum of challenge problems related to insider threat detection. We present in general terms the motivation for and description of our MIEM solution. We then present the results of applying MIEM across a range of challenge problems, with a detailed illustration for one challenge problem. Finally, we discuss the science and promise of IEM and MIEM, including the applicability of MIEM to a spectrum of inference domains. This article is categorized under: Technologies > Machine Learning Algorithmic Development > Ensemble Methods Technologies > Prediction
Common inductive inference problems
[ Normal View | Magnified View ]
(a) Histogram showing the distribution of the 60% certainty interval sizes for all questions in all challenge problems (CPs). (b) The median (black), maximum (orange), and minimum (blue) 60% certainty interval size for each CP
[ Normal View | Magnified View ]
Histograms showing variability of evaluation metrics across all challenge problems (CPs). Dashed vertical line indicates sample mean value,
[ Normal View | Magnified View ]
Multimodeling results for challenge problem (CP) 16
[ Normal View | Magnified View ]
Summary of evaluation metrics for challenge problems (CPs) 1–14 for the baseline model, competitors 1, competitors 2, and our team at IDI. The blue bars reference the CPs where the competitors were asked to identify behaviors (CPs 1–9), where as the orange bars reference the CPs where the competitors were asked to evaluate the the system (CPs 10–14). The baseline model used a Bayesian network to develop a reference point for evaluating the competitors. For the mean square error and Interval Scoring Rule, the target for the competitors was to be below 50% of the baseline (as indicated by the orange and blue dashed lines). For the Certainty Interval Calibration, the target was to be within 0.6 ± 0.2 (as indicated by the black solid and dashed lines)
[ Normal View | Magnified View ]
Multimodeling interaction and comparison
[ Normal View | Magnified View ]
Detailed schematic showing multimodeling inference enterprise modeling (MIEM) model
[ Normal View | Magnified View ]
Multimodeling inference enterprise modeling (MIEM) meta‐model
[ Normal View | Magnified View ]

Browse by Topic

Technologies > Prediction
Algorithmic Development > Ensemble Methods
Technologies > Machine Learning

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts