Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Data Mining Knowl Discov
Impact Factor: 2.541

A comprehensive survey of error measures for evaluating binary decision making in data science

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Binary decision making is a topic of great interest for many fields, including biomedical science, economics, management, politics, medicine, natural science and social science, and much effort has been spent for developing novel computational methods to address problems arising in the aforementioned fields. However, in order to evaluate the effectiveness of any prediction method for binary decision making, the choice of the most appropriate error measures is of paramount importance. Due to the variety of error measures available, the evaluation process of binary decision making can be a complex task. The main objective of this study is to provide a comprehensive survey of error measures for evaluating the outcome of binary decision making applicable to many data‐driven fields. This article is categorized under: Fundamental Concepts of Data and Knowledge > Key Design Issues in Data Mining Technologies > Prediction Algorithmic Development > Statistics
A summary of a binary decision making in the form of a contingency table
[ Normal View | Magnified View ]
Application field specific to relative usage frequency of error measures. The frequency numbers have been obtained from Scopus
[ Normal View | Magnified View ]
Comparison of error measures with respect to their relative usage frequency. The frequency numbers have been obtained from Google Scholar
[ Normal View | Magnified View ]
Outcome of three binary decision making methods. The error measures are evaluated for T = 100 samples
[ Normal View | Magnified View ]
Dashboard view of seven error measures according to the error model shown in Figure
[ Normal View | Magnified View ]
Visualization of the assumed error model that defines the proportions of the four fundamental errors
[ Normal View | Magnified View ]
In the first row two examples for ROC curves are shown resulting from two logistic regression analyses. The second row shows results for Figure 9a for the distance to the ROC curve and the Youden‐Index leading to optimal thresholds
[ Normal View | Magnified View ]
Behavior of the MCC according to the specificity value and the relationship between MCC and ACC, TPR and TNR. The used values for prevalence and sensitivity are mentioned in the figures
[ Normal View | Magnified View ]
Left: false negative rate. Right: false positive rate
[ Normal View | Magnified View ]
Left: false discovery rate. Right: false omission rate
[ Normal View | Magnified View ]
Behavior of the F‐score in dependence on the parameter β. The used values for PPV and sensitivity are mentioned in the figures
[ Normal View | Magnified View ]
Left: positive predictive value. Right: negative predictive value
[ Normal View | Magnified View ]
Left: true positive rate. Right: true negative rate
[ Normal View | Magnified View ]
Overview of different error measures for binary decision making
[ Normal View | Magnified View ]

Browse by Topic

Technologies > Prediction
Algorithmic Development > Statistics
Fundamental Concepts of Data and Knowledge > Key Design Issues in Data Mining

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts