Introduction To Machine Learning

This is an introductory tutorial on machine learning using the scikit-learn and Keras Python packages. Prerequisites are minimal; chiefly, I assume that the reader has a little bit of prior programming experience—preferably in Python. A passing familiarity with basic inferential statistical methods (primarily linear regression) is also helpful, but isn’t essential. Some of the material in thi tutorial is borrowed from Jake Vanderplas’s excellent scikit-learn tutorial. The main differences between the present tutorial and most others out there are that (a) this tutorial is more verbose than most (i.e., the emphasis is on conceptual understanding rather than just on learning the scikit-learn API), and (b) most of the examples are drawn from more unique datasets and contain application to real world examples.

Software

All of the code in this tutorial is written in Python. There is nothing intrinsically special about Python in the machine learning context; in principle, all of the examples and simulations in these pages could have been written in other languages (R, Matlab, etc.). Indeed, there are plenty of machine learning tutorials out there written in other languages.That said, Python does have a number of practical advantages over other languages. Chief among these is the fact that it’s currently the most widely used language in the data science and machine learning community. This means there are exceptional tools written in Python for virtually every domain of machine learning. Exhibit A is the scikit-learn package for machine learning. Scikit-learn is the world’s most widely used machine learning, and some of the reasons for its popularity will hopefully soon become clear. Scikit-learn is itself built on the numpy numerical computing library, which we’ll also use fairly regular. Exihbit B is the Keras package for machine learning. Keras is a high-level neural networks API, written in Python and capable of running on top of TensorFlow. It was developed with a focus on enabling fast experimentation.

The scikit-learn package

Now that we know what machine learning is, let’s turn to the scikit-learn package. Scikit-learn is the most widely-used machine learning package in Python (and probably the most widely-used ML package, period). Its popularity stems from its simple, elegant API, stellar documentation, and comprehensive support for many of the most widely used machine learning algorithms (the main exception being deep learning, which we will use Keras for). Scikit-learn provides well-organized, high-quality tools for virtually all aspects of the typical machine learning workflow, including data loading and preprocessing, feature extraction and feature selection, dimensionality reduction, model selection and evaluation, and so on.

Python Setup

First you will need to install Python, depending on what operating system you are using there are different approaches.

Mac

  1. Install Homebrew by opening a Terminal window and pasting the following line.
/usr/bin/ruby -e $(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)
  1. Homebrew asks you to enter your password so it can finalize the installation. Enter your user account password and hit enter
  2. Once Homebrew has finished installing, return to your terminal and run the following command:
brew install python3

Windows

  1. You will need to download the windows python installer.
  2. Underneath the heading at the top that says Python Releases for Windows, click on the link for the Latest Python 3 Release - Python 3.x.x
  3. Scroll to the bottom and select either Windows x86-64 executable installer for 64-bit or Windows x86 executable installer for 32-bit
  4. Install by double-clicking on the downloaded file.

Linux

  1. Open a terminal window and run the following commands:
sudo apt-get install python3.6
sudo apt install python3-pip

Python Interactor

Several Python integrated development environments (IDE) exist to make writting Python code easier. The one I use the most, and I highly recommend, is Spyder IDE. Spyder is a powerful scientific environment written in Python, for Python, and designed by and for scientists, engineers and data analysts. It offers a unique combination of the advanced editing, analysis, debugging, and profiling functionality of a comprehensive development tool with the data exploration, interactive execution, deep inspection, and beautiful visualization capabilities of a scientific package.

Once you have installed Python, installing Spyder is straight forward. You will need to open a terminal or command prompt and type the following

pip install spyder
Numpy
NumPy (or Numpy) is a Linear Algebra Library for Python, the reason it is so important for Data Science with Python is that almost all of the libraries in the PyData Ecosystem rely on NumPy as one of their main building blocks.
Last updated on Sep 30, 2021
Advanced Matplotlib
In this tutorial we cover some more advanced topics which you won’t usually use as often. Logarithmic scale It is also possible to set a logarithmic scale for one or both axes.
Last updated on Sep 30, 2021
Data Types
Data Types We often begin exploratory data analysis by examining the types of data that occur in a table. Although there are multiple ways of categorizing data types, in this book we discuss three broad types of data:
Last updated on Sep 30, 2021
Input Variables
Having some features we can use to make predictions is great, but we also need to know what we’re trying to predict! Conventionally, the target of the prediction process is a vector of scores usually labeled y.
Last updated on Sep 30, 2021
Pandas
Pandas is a newer package built on top of NumPy, and provides an efficient implementation of a DataFrame. DataFrames are essentially multidimensional arrays with attached row and column labels, and often with heterogeneous types and/or missing data.
Last updated on Sep 30, 2021
Data Transformations
Transformations of predictor variables may be needed for several reasons. Some modeling techniques may have strict requirements, such as the predictors having a common scale. In other cases, creating a good model may be difficult due to specific characteristics of the data (e.
Last updated on Sep 30, 2021
Distribution Plots
Let’s discuss some plots that allow us to visualize the distribution of a data set. These plots are: distplot jointplot pairplot rugplot kdeplot Imports import seaborn as sns %matplotlib inline Data Seaborn comes with built-in data sets!
Last updated on Sep 30, 2021
Categorical Data Plots
Now let’s discuss using seaborn to plot categorical data! There are a few main plot types for this: factorplot boxplot violinplot stripplot swarmplot barplot countplot Let’s go through examples of each!
Last updated on Sep 30, 2021
Matrix Plots
Matrix plots allow you to plot data as color-encoded matrices and can also be used to indicate clusters within the data (later in the machine learning section we will learn how to formally cluster data).
Last updated on Sep 30, 2021
Grids Plots
Grids are general types of plots that allow you to map plot types to rows and columns of a grid, this helps you create similar plots separated by features. import seaborn as sns import matplotlib.
Last updated on Sep 30, 2021