Petroleum reservoir characterization is one of the most difficult and challenging tasks of the exporation of petroleum industry and usually a long and costly procedure. This paper proposes a novel kind of patterns for the classification over quantitative well logging data, which is called MOUCLAS (MOUntain function based CLASsification) Patterns, based on the concept of the fuzzy set membership function which gives the new approach a solid mathematical foundation and compact mathematical description of classifiers. It integrates classification, clustering and association rules mining to identify interesting knowledge in the well logging database. The aim of the study is the use of MOUCLASS patterns to interpret the pay zones from well logging data for the purpose of reservoir characterization. This approach is better than conventional techniques for well logging interpretation that require a precise understanding of the relation between the well logging data and the underlying property of interest.