Application of Decision Trees for Candidate Well Selection for Geological and Technical Measures

This paper examines the possibility of using Data Mining methods for analyzing the efficiency of geological and technical measures and selecting candidate wells for hydraulic fracturing. A three-step approach is proposed. It includes the classification of already performed hydraulic fracturings, the identification of parameters that significantly affect the effectiveness of hydraulic fracturing, and the well classification model training. A search for the meaningful parameters is based on the Student's distribution. Classification decision trees is a one of the machine learning methods that is used for forecasting the effectiveness of geological and technical measures. Decision trees allow analysis in two directions. First, the trained trees allow obtaining new knowledge, for instance, the reasons of hydraulic fracturing inefficiency. Secondly, they allow selecting a candidate wells for effective hydraulic fracturing.