Currently Data Mining methods acquire popularity as a tool for analyzing volumetric information; such methods are based on both classical principles of exploratory data analysis and modern ones, including neural networks. Due to their high practical significance, Data Mining methods make it possible to compactly describe data, understand their structure, and classification, discover patterns in a chaos of random phenomena. The article is devoted to the issue of using detailed statistical analysis of some characteristics of directional wells for the purpose of predictive assessment of the oil potential of new wells or sidetracks. According to the authors, it is possible to increase the predictive potential of Data Mining methods only when the production characteristics of wells are predicted without using of production data, relying only on geological and technological parameters. The authors solve regression and classification problems at the intersection of exploratory data analysis and Data Mining methods, assess the forecasting accuracy of the algorithms used on several samples, and also demonstrate the statistically substantiated influence of geological and technological parameters on the production characteristics of wells. The performed studies confirm the adequate forecasting ability of geo-statistical models in conditions of data limitations. Working with categorized variables under the conditions of a classifier, an approach has been proposed that allows reducing significantly the probability of prediction error. Matrices of the influence of parameters that are valid for all directional wells of the object under consideration are derived. The importance and indispensability of data analysis technologies is indicated that allow obtaining data from 2D and 3D geological models, and, as a result, assessing the production potential and efficiency of well placement already at the modeling stage. Thus, it is proposed to introduce data analysis technologies into the field development process that are capable of describing volumetric data, identifying patterns, conducting classification and forecasting under conditions of uncertainty.
References
1. StatSoft.Ink, URL: https://www.statsoft.ru/home/textbook/default.htm
2. STATISTICA. Ofitsial’noe rukovodstvo (STATISTICA. Official Guide), Vol. 3, 2007.
3. Neyronnye seti. STATISTICA Neural Networks: Metodologiya i tekhnologii sovremennogo analiza dannykh (Neural networks. STATISTICA Neural Networks: Methodology and technologies of modern data analysis): edited by Borovikov V.P., Moscow: Goryachaya liniya – Telekom Publ., 2008, 392 p.
4. Borovikov V.P., STATISTICA. Iskusstvo analiza
