Application of data mining for handling applied problems of petroleum engineering

Abstract

The fields of application of machine learning in the oil industry are actively expanding. Despite this, there are currently no convenient and simple tools that allow you to use machine learning methods to solve applied problems without special programming skills. The purpose of this work is to create a program that will allow to carry out data mining using machine learning algorithms and solve common problems associated with the analysis and construction of predictive models. Created an algorithm to implementation typical stages of the data analysis process (detection of abnormal values, filling the skipped values, smoothing the time series, reducing the dimension of the original feature space) and build a predictive models. Test examples showed that the developed program allows to construct a predictive models, as well as the search for significant features, which is applicable both for the construction of surrogate models for the optimization of oilfield development and for an analysis of hydrodynamic connectivity of wells.