Oil and Gas

Participants will learn to:

1. Discuss the subsurface applications of machine learning.
2. Perform data clean-up, outlier detection and handling, and visualization.
3. Apply a structured approach to unsupervised clustering and cluster evaluation.
4. Perform supervised methods for classification and regression.
5. Critically evaluate the various machine learning algorithms.

Introduction

• Terminology related to data-driven modelling
• Key concepts related to supervised, unsupervised, and reinforcement learning
• A discussion of use cases relevant to the engineering and geoscience disciplines

Data Processing

• Basic probability, univariate, and multivariate statistics
• Exploratory data analysis, visualization and plotting multivariate data
• Data imputation for numerical and categorical data
• Outlier handling

Machine Learning Workflows

• Transformations: Scaling, dimensionality reduction
• Unsupervised learning: K-means and K-Prototype clustering, variants of Hierarchical Clustering, DBSCAN, Gaussian Mixture Models
• Supervised learning:
• Model Evaluation, Cross Validation, and Parameter Selection
• Classification: Decision Trees, Random Forests, Gradient Boosting, K-Nearest Neighbours, Support Vector Machines, Neural Networks
• Regression: K-Nearest Neighbours, Neural Networks, Decision Trees and Random Forests, Multilinear Regression
• Feature selection, ranking, and elimination
• Use cases in the E&P industry

The course is designed for geoscientists and petroleum engineers with little to no prior experience in machine learning, seeking to incorporate machine learning tools in their E&P workflows. It is expected that participants have a working knowledge of programming and Python. A Python self-assessment rubric is provided in advance of the course; If participants need a Python refresher, a series of pre-course self-paced videos and exercises will prepare participants with the necessary background for the machine learning course.