Marcellus Shale Pattern Recognition Study Weighs Parameters' Impact MORGANTOWN, W.V.-To better understand the inherent complexities of their behavior, analytical, numerical and statistical analyses have been applied to large, multivariable datasets from shale assets with different degrees of success. Pattern recognition technologies with roots in machine learning have proven capable of extracting useful information from large datasets. Extensively used in many industries, they also can be applied in multivariable datasets from shale assets to extract understandable structure in the data. This article presents a pattern recognition study of well locations and trajectories, reservoir characteristics, and completion, hydraulic fracturing and production parameters for a large number of horizontal Marcellus Shale wells. The shale assets' dataset is so complex that conventional statistical analysis cannot reveal understandable trends and patterns. On the other hand, advanced pattern recognition tools allow certain previously hidden patterns to emerge from the data with unmistakable trends. 52 THE AMERICAN OIL & GAS REPORTER