Book Image

R Data Mining

Book Image

R Data Mining

Overview of this book

R is widely used to leverage data mining techniques across many different industries, including finance, medicine, scientific research, and more. This book will empower you to produce and present impressive analyses from data, by selecting and implementing the appropriate data mining techniques in R. It will let you gain these powerful skills while immersing in a one of a kind data mining crime case, where you will be requested to help resolving a real fraud case affecting a commercial company, by the mean of both basic and advanced data mining techniques. While moving along the plot of the story you will effectively learn and practice on real data the various R packages commonly employed for this kind of tasks. You will also get the chance of apply some of the most popular and effective data mining models and algos, from the basic multiple linear regression to the most advanced Support Vector Machines. Unlike other data mining learning instruments, this book will effectively expose you the theory behind these models, their relevant assumptions and when they can be applied to the data you are facing. By the end of the book you will hold a new and powerful toolbox of instruments, exactly knowing when and how to employ each of them to solve your data mining problems and get the most out of your data. Finally, to let you maximize the exposure to the concepts described and the learning process, the book comes packed with a reproducible bundle of commented R scripts and a practical set of data mining models cheat sheets.
Table of Contents (22 chapters)
Title Page
Credits
About the Author
About the Reviewers
www.PacktPub.com
Customer Feedback
Preface
14
Epilogue

Measuring performance in regression models


Let's make it more technical and discuss the most common metrics available when dealing with regression models. First of all, let's recall what a regression is: we are trying to explain here a response variable with a set of explanatory variables. A reasonable model performance metric will therefore be one that summarizes how well our model is able to explain the explanatory variable itself. 

It is no surprise that the most popular regression model metrics are both able to explain this:

  • Mean squared error
  • R-squared

Both of them are based on the concept of error, which we already encountered when dealing with model coefficient estimation. We defined as error for a given record of the estimation dataset the difference between the actual value of the response variable and the value of the response value we estimate with our model:

We also called this residual, and employed it to test some of the most relevant assumptions regarding linear regression models...