Hands-On Machine Learning with Microsoft Excel 2019

Build complete data analysis flows, from data collection to visualization

Julio Cesar Rodriguez Martino

• Publisher: Packt Publishing
• Year: 2019
• Print ISBN: 9781789345377
• eBook ISBN: 9781789345124
• Edition: 1
• Topic: Computer Science
• Subtopic: Artificial Intelligence (AI) & Semantics
• Index: Computer Science

A practical guide to getting the most out of Excel, using it for data preparation, applying machine learning models (including cloud services) and understanding the outcome of the data analysis.

Key Features

• Use Microsoft's product Excel to build advanced forecasting models using varied examples
• Cover range of machine learning tasks such as data mining, data analytics, smart visualization, and more
• Derive data-driven techniques using Excel plugins and APIs without much code required

Book Description

We have made huge progress in teaching computers to perform difficult tasks, especially those that are repetitive and time-consuming for humans. Excel users, of all levels, can feel left behind by this innovation wave. The truth is that a large amount of the work needed to develop and use a machine learning model can be done in Excel.

The book starts by giving a general introduction to machine learning, making every concept clear and understandable. Then, it shows every step of a machine learning project, from data collection, reading from different data sources, developing models, and visualizing the results using Excel features and offerings. In every chapter, there are several examples and hands-on exercises that will show the reader how to combine Excel functions, add-ins, and connections to databases and to cloud services to reach the desired goal: building a full data analysis flow. Different machine learning models are shown, tailored to the type of data to be analyzed.

At the end of the book, the reader is presented with some advanced use cases using Automated Machine Learning, and artificial neural network, which simplifies the analysis task and represents the future of machine learning.

What you will learn

• Use Excel to preview and cleanse datasets
• Understand correlations between variables and optimize the input to machine learning models
• Use and evaluate different machine learning models from Excel
• Understand the use of different visualizations
• Learn the basic concepts and calculations to understand how artificial neural networks work
• Learn how to connect Excel to the Microsoft Azure cloud
• Get beyond proof of concepts and build fully functional data analysis flows

Who this book is for

This book is for data analysis, machine learning enthusiasts, project managers, and someone who doesn't want to code much for performing core tasks of machine learning. Each example will help you perform end-to-end smart analytics. Working knowledge of Excel is required.

Section 1: Machine Learning Basics

The objective of part 1 is to introduce the reader to machine learning and the different types of models used. It will cover supervised and unsupervised learning, the principal division within machine learning. Within these aspects, the difference between regression (continuous target variable) and classification (discrete target variable) will be demonstrated. All points are explained by means of hands-on examples.

This section comprises the following chapters:

• Chapter 1, Implementing Machine Learning Algorithms
• Chapter 2, Hands-On Examples of Machine Learning Models

Implementing Machine Learning Algorithms

Learning has been a matter of study for many years. How human beings acquire new knowledge, from basic survival skills to advanced abstract subjects, is difficult to understand and reproduce in the computer world. Machines learn by comparing examples and by finding similarities in them.

The easiest way for a machine (and also for a human being) to learn is to simplify the problem that needs to be solved. A simplified version of reality, called a model, is useful for this task. Some of the relevant issues to be studied are the minimum number of samples, underfitting and overfitting, relevant features, and how well a model can learn. Different types of target variables require different algorithms.

In this chapter, the following topics will be covered:

• Understanding learning and models
• Focusing on model features
• Studying machine learning models in practice
• Evaluating models

Technical requirements

There are no technical requirements for this chapter, since it is introductory. The data shown in the sections should be input into an Excel spreadsheet in order to be able to follow the examples.

Understanding learning and models

The way that humans learn has been studied for many decades now. There are a handful of psychological theories that try to explain how we acquire knowledge, use it, and generalize it in order to apply what we know to completely new scenarios. Taking one step back, we could ask ourselves: what does it mean to learn? We could say that, once we learn something, we are able to repeat it in a more or less detailed way. In reality, learning implies much more than just copying a behavior or memorizing a piece of poetry. In fact, we understand what we learn and are able to generalize that knowledge, which helps us to react correctly to new people, places, and situations.

The need to create a machine that somehow mimics our human behavior and intelligence has been desired for a very long time. Hundreds of years ago, kings were amazed by chess-playing machines, musical instruments that did not require a human player, and mysterious boxes that answered all kinds of questions. These, many times fake, inventions show that one of the greatest dreams of humans is to create an intelligent being, which is able to replicate easy or difficult tasks that are usually performed by people, even when intelligence is an elusive and not easily-defined thing.

Many years have passed, and technology has evolved in such a way that we can now create machines that think, or at least seem to. In fact, most of the systems that we call intelligent are just able to perform repetitive tasks or react to external inputs according to whatever we have showed them by way of example. As we progress through the chapter, we will see that some of the defined characteristics of human learning and intelligence are already part of modern machine learning systems and some are still the subject of science fiction novels.

By definition, machine learning means to teach a machine or an algorithm to perform tasks. We have been doing this for many years now – it is called programming. We give a computer a set of instructions, the order in which they should be executed, and a number of options of how to react to a limited number of inputs. If the input is not known, or if we ask the computer to do something that is not contained in the program, then it will fail, showing an error. The difference between this traditional paradigm and machine learning is that we will never tell the computer exactly what to do. We will either let it discover patterns or show it samples of what we want. We will use programming, of course, but just to define algorithms that learn in the sense that was described previously. From finding the straight line that better represents a set of points to driving a car, everything a machine can do is learned in this way.

As babies, we start exploring the world around us. Since we are too young to understand words or examples, we basically experience the world through our senses. We learn the difference between hard and soft, rough and smooth, hot and cold. We can call for attention when we need something, and we can even gain an understanding of the patience levels of our parents and pets. In most cases, nobody sits next to us to explain what we see, hear, feel, taste, and smell. This is an example of what we call unsupervised learning.

In unsupervised learning, the training data is "unlabeled". Without our help or intervention, the algorithm/s (or program/s) will find the required connections or unsuspected patterns in the data and learn the details and properties of the dataset.

Later, as we grow up, we understand words and start naming things. Our parents tell us when we see a dog or a cat, we learn our names and theirs, and we learn to identify our toys from among other children's toys (and fight over them). Without even realizing it, we relate some characteristics of objects, animals, and people to their names. These are examples of what is known as supervised learning. In the case of a computer, the algorithm is shown as a set of variables that are representative of the properties of the problem and then it learns how these features relate to the name of the label.

Science has shown us the immense complexity of the world that surrounds us. Every branch of scientific knowl...

Table of contents

• Title Page
• Copyright and Credits
• Dedication
• About Packt
• Contributors
• Preface
• Section 1: Machine Learning Basics
• Implementing Machine Learning Algorithms
• Hands-On Examples of Machine Learning Models
• Section 2: Data Collection and Preparation
• Importing Data into Excel from Different Data Sources
• Data Cleansing and Preliminary Data Analysis
• Correlations and the Importance of Variables
• Section 3: Analytics and Machine Learning Models
• Data Mining Models in Excel Hands-On Examples
• Implementing Time Series
• Section 4: Data Visualization and Advanced Machine Learning
• Visualizing Data in Diagrams, Histograms, and Maps
• Artificial Neural Networks
• Azure and Excel - Machine Learning in the Cloud
• The Future of Machine Learning
• Assessment