Polytopic Vector Analysis - Non-Constant Sum (PVA-NCS): Confidence and Risk Assessment

William Full¹ and Anthony C. Gary^2
1 Tramontane Inc, Salt Lake City, UT
² Energy & Geoscience Institute, Salt Lake City, UT

There are many possible flaws in routine geologic data that can obscure valuable patterns. These include noise, inconsistencies in precision, uneven accuracy, incorrect processing assumptions and variables that do not contain relevant information. Considering these many sources of data flaws and their inherent complexities, it is important that an analyst have some indication regarding the confidence and risk associated with any analytical data mining solution.

A comprehensive data mining technique must: 1) cope with these flaws and differences in data structures; 2) define a confidence associated with using a particular technique; 3) not impose a structure or force a solution; 4) define “sub-systems” in large data arrays in an unbiased manner; 5) give useful solutions that incorporate risk assessment principles, and 6) produce solutions that can be readily incorporated with other data, such as seismic, potential methods, petrophysical information, reservoir characteristics, and so forth.

Polytopic Vector Analysis – Non-constant Sum (PVA-NCS) is a technique developed over the last twenty-five years that meets the above requirements. PVA-NCS is a logical extension of the Polytopic Vector Analysis (PVA) approach that has been extensively used under a variety of names since the 1960's. The PVA method has been proven with varied geological data sets, such as wireline logs, geochemical and biostratigraphic data to name a few. A major advantage of the PVA-NCS approach in data mining is that and it produces results in the original data metric that facilitates direct interpretation of the original data and a priori knowledge can be incorporated wherever appropriate. Applications of PVA-NCS data mining will be presented that demonstrate the concept of confidence and its relationship to dynamic risk assessment.