ABSTRACT: New Tools for Discovery from Old Databases

James P. Brown

Very large quantities of information have been accumulated as a result of petroleum exploration and the practice of petroleum geology. New and more powerful methods to build and analyze databases have been developed. The new tools must be tested, and, as quickly as possible, combined with traditional methods to take full advantage of currently limited funds in the search for new and extended hydrocarbon reserves.

A recommended combined sequence is (1) database validating, (2) category separating, (3) machine learning, (4) graphic modeling, (5) database filtering, and (6) regression for predicting.

To illustrate this procedure, a database from the Railroad Commission of Texas has been analyzed. Clusters of information have been identified to prevent apples and oranges problems from obscuring the conclusions.

Artificial intelligence has checked the database for potentially invalid entries and has identified rules governing the relationship between factors, which can be numeric or nonnumeric (words), or both. Graphic 3-Dimensional modeling has clarified relationships. Database filtering has physically separated the integral parts of the database, which can then be run through the sequence again, increasing the precision. Finally, regressions have been run on separated clusters giving equations, which can be used with confidence in making predictions.

Advances in computer systems encourage the learning of much more from past records, and reduce the danger of prejudiced decisions. Soon, there will be giant strides beyond current capabilities to the advantage of those who are ready for them.

AAPG Search and Discovery Article #91003©1990 AAPG Annual Convention, San Francisco, California, June 3-6, 1990