Thankfully, there is a commonly acceptable format in which well logs are expected to be.

LAS is an industry-standard file format used in all oil-and-gas and water well industries to log and store well log information and data. A single LAS file can only contain data for one well. But in that one well, it can contain any number of datasets (called curves). Common curves found in an LAS file may include natural gamma, travel time, resistivity logs, and other possible information.

Wow! Although not rocket science, the data is not a simple relational table. Preliminary work for this exercise will be to better understand the specifics of the data provided.

In this chapter, our goal is to implement a machine learning algorithm in Python using scikit-learn, one of the most popular machine learning tools for Python, based upon a sample well drilling log dataset for the task of training a classifier to distinguish between different types of lithofacies.

Suppose that we are told that the training log dataset was created from sample logs, and based upon research defining eight different lithofacies, along with various log measurements, such as gamma-ray, neutron porosity, photoelectric factor (PeF), and resistivity.

We also know that in this file, we'll have six lithofacies data points (GCR, NPHI, PE, PEF, ILD, and ILM), along with an ID and the lithofacies type.

The following screenshot shows a portion of the top section of an actual well logging file:

The following screenshot is a peek at a portion of our file, which we will use for training our machine learning model. This file excludes the top LAS-formatted section headers and is simply a continuous list of curves or well-logging measurements: