Machine learning workflow

In the next section, we'll go through the basics of time-series and machine learning.

Machine learning mostly deals with numerical data that is in tabular form as a matrix of size . The layout is generally in a way that each row represents an observation, and each column represents a feature.

In time-series problems, the column related to time doesn't necessarily serve as a feature, but rather as an index to slice and order the dataset. Time columns can, however, be transformed into features, as we'll see in Chapter 3, Preprocessing time-series.

Each observation is described by a vector of M features. Although a few machine learning algorithms can deal with non-numerical data internally, typically, each feature is either numerical or gets converted to numbers before feeding it into a machine learning algorithm. An example of a conversion is representing Male as 0 and Female as 1. Put simply, each feature can be defined as follows...