Low false positive learning with support vector machines

In automated computer-based diagnosis systems, falsely determining that a case is normal is much more serious than falsely determining that the case is abnormal, especially if the system is being used for triage of patients. This is an example where false positives and false negatives are not equally bad. In this paper, the authors solve the problem proposing a new method: Risk Area SVM (RA-SVM).

The solution introduces a second classifier, which analyzes the k nearest neighbors (k-NN) within a defined risk area and classifies the data in this region as positive only if all its k-nearest neighbors are also positive. The authors compare the solution against the state-of-the-art methods for low false positive classification using 33 standard datasets. The obtained results demonstrate that RA-SVM has better performance in the vast majority of the cases.

An example of the RA-SVM, with the risk area. In this case, we have k = 2. The five testing points inside the risk area have their classes defined by the class of their 2-nearest neighbors. The positive samples are represented by circles and the negative samples by squares. The samples that are classified as positive by the Risk Area SVM are highlighted in green while those that are classified as negative are in blue.

An example of the RA-SVM, with the risk area. In this case, we have k = 2. The five testing points inside the risk area have their classes defined by the class of their 2-nearest neighbors. The positive samples are represented by circles and the negative samples by squares. The samples that are classified as positive by the Risk Area SVM are highlighted in green while those that are classified as negative are in blue.


Daniel Moraes, Jacques Wainer, Anderson Rocha, Low false positive learning with support vector machines, Journal of Visual Communication and Image Representation, Volume 38, July 2016, Pages 340-350, ISSN 1047-3203, doi:10.1016/j.jvcir.2016.03.007.

Advertisements
This entry was posted in blog, publications and tagged , , , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s