Attribute importance measure for parametric multivariate modeling

摘要

Parametric Bayesian Updating method is described for attribute importance measures. AIM is a quantitative measure of the relative importance of multiple attributes used in a Bayesian Updating parametric multivariate modeling approach. The AIM is a function of the correlation coefficients between the attributes themselves and the attributes with collocated target variable. The importance calculation is straightforward, repeatable, and objective.

主权项

1. A method comprising:
a) making a multivariate Gaussian assumption with a normal score transformation for any events within a subsurface formation; b) executing a simple kriging to estimate a prior likelihood mean (ySK) and variance (δSK2) for the events; c) calculating, by a processor, a matrix of correlation coefficients (ρn) from all collocated combinations of normal scored event pairs; d) executing a simple kriging (λn) with a calculated model of multivariate correlation to estimate a likelihood mean (yL) and variance (δL2); e) calculating an updated Gaussian Probability Distribution Function (PDF) by combining the prior and likelihood estimates according to the normal score transformation (a); f) transforming the updated Gaussian PDF N(yU, δU2) back to real units using the normal score transformation (a); g) calculating one or more attribute importance (In) using equationIn=λn·ρn∑n=1N⁢⁢λn·ρn,n=1,…⁢⁢N;  and h) obtaining formation measurements for attributes with an attribute importance (I) over exceedance probability of 50% to parameterize a final multivariate model that predicts porosity, permeability, water saturation, or net thickness.