Learning to Quantify

Learning to quantify is the task of training estimators of class proportions in unlabeled data by supervised learning, which is different from classification. This open access book provides an introduction to the topic, covering supervised learning methods, evaluation measures, and applications in various fields. It is suitable for researchers, data scientists, and PhD students who want to learn about quantification and apply it to their work.

Format: Paperback / softback
Length: 137 pages
Publication date: 17 March 2023
Publisher: Springer International Publishing AG

This comprehensive open-access book offers an in-depth introduction and comprehensive overview of the field of learning to quantify, also known as "quantification." It delves into the task of training estimators of class proportions in unlabeled data through supervised learning. In the realm of data science, learning to quantify stands as a distinct domain, closely related to classification but distinct in its approach. While it may seem intuitive to estimate class proportions by simply classifying all data and counting the labels assigned by a classifier, this method often yields inaccurate and biased estimates.

The book provides a comprehensive exploration of learning to quantify by examining the various supervised learning methods employed in this process. It also discusses the essential evaluation measures and evaluation protocols that should be employed to assess the quality of the predictions generated. Furthermore, the book highlights the numerous fields of human activity where the application of quantification techniques can yield superior results compared to the naïve use of classification techniques. It delves into advanced topics in quantification research, making it an invaluable resource for researchers, data scientists, and Ph.D. students seeking to stay abreast of the latest developments in this field.

Moreover, this book is particularly suited to researchers who wish to apply data science technologies to fields of human activity that focus on aggregate (or "macro") data rather than individual (or "micro") data. These fields include social sciences, political science, epidemiology, market research, and more. By leveraging the power of quantification, researchers can gain insights into complex systems, make informed decisions, and drive positive change in a wide range of domains.

In summary, this open-access book serves as a comprehensive guide to learning to quantify, providing a solid foundation for researchers, data scientists, and students interested in this field. Its in-depth exploration of supervised learning methods, evaluation measures, and application domains makes it an essential resource for anyone seeking to advance their understanding and expertise in quantification.

Weight: 250g
Dimension: 235 x 155 (mm)
ISBN-13: 9783031204661
Edition number: 1st ed. 2023