Machine Learning for Data Mining

Improve your data mining capabilities with advanced predictive modeling

Jesus Salcedo

• Publisher: Packt Publishing
• Year: 2019
• ISBN:
  9781838821555
• Edition: 1
• Topic: Ciencia de la computación
• Subtopic: Minería de datos

Get efficient in performing data mining and machine learning using IBM SPSS Modeler

Key Features

• Learn how to apply machine learning techniques in the field of data science
• Understand when to use different data mining techniques, how to set up different analyses, and how to interpret the results
• A step-by-step approach to improving model development and performance

Book Description

Machine learning (ML) combined with data mining can give you amazing results in your data mining work by empowering you with several ways to look at data. This book will help you improve your data mining techniques by using smart modeling techniques.

This book will teach you how to implement ML algorithms and techniques in your data mining work. It will enable you to pair the best algorithms with the right tools and processes. You will learn how to identify patterns and make predictions with minimal human intervention. You will build different types of ML models, such as the neural network, the Support Vector Machines (SVMs), and the Decision tree. You will see how all of these models works and what kind of data in the dataset they are suited for. You will learn how to combine the results of different models in order to improve accuracy. Topics such as removing noise and handling errors will give you an added edge in model building and optimization.

By the end of this book, you will be able to build predictive models and extract information of interest from the dataset

What you will learn

• Hone your model-building skills and create the most accurate models
• Understand how predictive machine learning models work
• Prepare your data to acquire the best possible results
• Combine models in order to suit the requirements of different types of data
• Analyze single and multiple models and understand their combined results
• Derive worthwhile insights from your data using histograms and graphs

Who this book is for

If you are a data scientist, data analyst, and data mining professional and are keen to achieve a 30% higher salary by adding machine learning to your skillset, then this is the ideal book for you. You will learn to apply machine learning techniques to various data mining challenges. No prior knowledge of machine learning is assumed.

Understanding Models

In this chapter, we're going to look into general model interpretation. We will have a look at the different types of predictive models. Then we will interpret some machine learning models using various techniques.

We will be covering the following topics in this chapter:

• Types of models
• Using graphs to interpret machine learning models
• Using statistics to interpret machine learning models
• Using decision trees to interpret machine learning models

Models

There are three different types of predictive models:

• Statistical models
• Decision tree models
• Machine learning models

Statistical models

The first thing that the statistical models identify is which predictors are most important in a model. The statistical models also create an equation that allows you to make predictions. For example, as we can see in the following screenshot, the coefficients that are part of the prediction equation have been highlighted:

The following screenshot highlights the equation to predict current salaries:

In the following screenshot, we can see that we take the coefficient for the variable beginning salary and we multiply it by the actual beginning salary:

Now, we take the coefficient for education level and we multiply it by a person's number of years of education:

We also need the person's age, which is highlighted in the following screenshot:

We multiply all the values with their respective coefficients, and finally, we add all the constants, which predicts what the person's salary is going to be. Now that's great. But in addition, the statistical models allow us to determine the effect of a one-unit increase in each predictor, and you can see the effect of this predictor on the different outcome variables.

So, for example, with education level, we see that it has a coefficient of 298. That tells us that a person's current salary is increasing by $298 for each additional year of education that the person has:

So you can really see the impact of each individual independent variable and how it ends up impacting the overall prediction.

Decision tree models

Just like statistical models, decision tree models help you identify which predictors are most important in the model. There are no equations, and we can't determine a one-unit impact and along with the effect that it has on an outcome variable. Instead we are going to create rules that make predictions by segmenting data in two mutually-exclusive categories.

So, for example, as you can see in the following screenshot, we have anybody with a variable Premier value of No:

We have anybody that was a customer for three years or fewer, as shown in the following screenshot:

And we have people whose estimated revenue was less than or equal to about 4,000,000, as shown in the following screenshot:

They happen to be located in another country, and we're predicting that they're going to churn, as shown in the following screenshot:

This rule was applied to 25 people, and the accuracy of that rule was 100%, as shown in the following screenshot. So, for all the customers who fit those criteria, we ended up losing them as customers 100% of the time.

Next, there's a second rule, and it's exactly the same as the first. The only difference is that the location is a National customer, and you will notice that we're still predicting the people who will be churned. That rule applies to 47 individuals, but the accuracy of that rule is only 66%:

Now, at first glance it might seem tha...

Table of contents

• Title Page
• Copyright and Credits
• Contributors
• About Packt
• Preface
• Introducing Machine Learning Predictive Models
• Getting Started with Machine Learning
• Understanding Models
• Improving Individual Models
• Advanced Ways of Improving Models
• Other Books You May Enjoy