In this course, the participants get access to codes and workflows in Python and they apply these pre-built software tools on the various types of the datasets. It is a hands-on course that allows participants to learn by assembling various pre-built programming modules to design interesting implementations of machine learning.
We’ll look at the realities of machine learning approach through the lens of applied field cases relevant to subsurface characterization and petroleum engineering.
Duration
1 day
Intended Audience
Intermediate Level
Prerequisites (Knowledge/Experience/Education required)
Basic Python Programming, Numerical Methods, Basic Statistics, Familiarity with basic machine learning concepts
This course will build on the entry-level content available in the SEG Shop.
Equipment/Software Requirements
The instructor will use Windows OS during the course. Participants will execute pre-built Python and modules/codes to understand various Machine Learning concepts and process various datasets. Participants will not be required to write codes. All software used for the course is open source, so participants should bring computers where they can install the open-source software and codes. Participants need at least 4GB of storage and 4GB RAM on their computer.
Objectives
- Participants will work on waveforms, images, depth-based and time-series data
- Participants will use Elastic Net, Support Vector Machine, Gradient Boosting, Neural Networks, Nearest Neighbor, Random forest, Mean-Shift, K-Means and Agglomerative clustering.
- Participants will be proficient in cross validation, hyperparameter optimization, feature elimination, feature selection, and model evaluation tasks.
- Participants can perform exploratory data analysis on datasets containing numerical and categorical data.
Course Outline
- Case Study #1 – Irreducible Saturation Estimation using Regressors
- Case Study #2 – Image Analysis using Feature Extraction and Random Forest
- Case Study #3 – Production Forecasting using Neural Networks and Gradient Boosting
- Case Study #4 – Waveform Analysis using Mean Shift and Agglomerative Clustering