Imbalanced classification with Python: choose better metrics, balance skewed classes, and apply cost-sensitive learning
No Preview Available
Citations
Altmetric:
Altmetric::
Files in this item
Associated Records
Authors
Editor
Date
2021
Source Title
Publisher
Machine Leaning Mastery
Melbourne Authors
Affiliation
Course Work
Document Type
Book
Subject Code
Citations
Access Status
Only available to University of Melbourne staff and students, login required
DOI
Access URL
Open Access URL
Open Access at PMC
Linked Resource URL
NHMRC Grant Code
ARC Grant Code
Description
1 online resource (1 PDF file (xvi, 446 pages) + 1 folder of data and code examples in text format).
System requirements: Adobe Reader and text file reader such as Notepad.
Classification predictive modeling involves assigning a class label to an example. It may be one of the most studied and used areas of machine learning. Nevertheless, the majority of the models used to learn from classi cation data and the metrics used to evaluate those models assume that the distribution of the examples across the class labels is equal. The eld is focused on the simplest form of classi cation problems, so-called balanced classification problems. When the examples across classes are imbalanced, many machine learning algorithms fail and metrics used to evaluate those models, such as classification accuracy, become dangerously misleading. Many problems have a skew in the class distribution. For example, numerous tasks where practitioners want to apply machine learning are examples of imbalanced classification, including fraud detection, churn prediction, medical diagnosis, and many more. In fact, it may be more common to have imbalanced classes than balanced classes. Thankfully, there is a small but rapidly growing eld of study dedicated to the problem of imbalanced classification. This field includes modifications to existing algorithms to make them useful for imbalanced classification, careful selection of performance metrics, and entirely new data preparation techniques and modeling algorithms. This book was carefully designed to help you bring the tools and techniques of imbalanced classification to your next project. The tutorials were designed to teach you these techniques the fastest and most e ective way that I know how: to learn by doing, with executable code that you can run to develop the intuitions required and that you can copy-and-paste into your project and immediately get a result.
Classification predictive modeling involves assigning a class label to an example. It may be one of the most studied and used areas of machine learning. Nevertheless, the majority of the models used to learn from classi cation data and the metrics used to evaluate those models assume that the distribution of the examples across the class labels is equal. The eld is focused on the simplest form of classi cation problems, so-called balanced classification problems. When the examples across classes are imbalanced, many machine learning algorithms fail and metrics used to evaluate those models, such as classification accuracy, become dangerously misleading. Many problems have a skew in the class distribution. For example, numerous tasks where practitioners want to apply machine learning are examples of imbalanced classification, including fraud detection, churn prediction, medical diagnosis, and many more. In fact, it may be more common to have imbalanced classes than balanced classes. Thankfully, there is a small but rapidly growing eld of study dedicated to the problem of imbalanced classification. This field includes modifications to existing algorithms to make them useful for imbalanced classification, careful selection of performance metrics, and entirely new data preparation techniques and modeling algorithms. This book was carefully designed to help you bring the tools and techniques of imbalanced classification to your next project. The tutorials were designed to teach you these techniques the fastest and most e ective way that I know how: to learn by doing, with executable code that you can run to develop the intuitions required and that you can copy-and-paste into your project and immediately get a result.
Abstract
Notes
Terms of Use
© 2021 Jason Brownlee. Reproduced with permission.
Keywords
Machine learning, Cost effectiveness, Artificial intelligence, Data processing, Machine learning technique