Autoencoder vs. Regression Neural Networks for Detecting Manipulated Wine Ratings

Titelangaben

Baumann, Michaela ; Baumann, Michael Heinrich:
Autoencoder vs. Regression Neural Networks for Detecting Manipulated Wine Ratings.
In: Mäkiö, Juho (Hrsg.): ICCGI 2022, The Seventeenth International Multi-Conference on Computing in the Global Information Technology. - Venice, Italy : IARIA , 2022 . - S. 7-13
ISBN 978-1-61208-972-0

Volltext

Link zum Volltext (externe URL):

Weitere URLs

Abstract

In this study, we analyze the ability of different (neural network based) detection methods to identify manipulated wine ratings for two “vinho verde” datasets. We find that autoencoders perform best on unmanipulated test data. However, regressions outperform autoencoders in terms of true/false positive rates on the manipulated test data in median. This is interesting, since autoencoders are generally used for outlier
detection. Furthermore, hyperparameter tuning via sequential accumulative selection is established.