As a guest user you are not logged in or recognized by your IP address. You have
access to the Front Matter, Abstracts, Author Index, Subject Index and the full
text of Open Access publications.
In this study, we address the problem of imbalanced data reducing machine learning performance. An analysis of the problem indicates that it is caused by differing amounts of data in each class and too few features in the minority class data. To address these issues, we propose a generative model that creates sufficient additional minority class data to eliminate the imbalance in the data amounts. This can also resolve the issue of a lack of minority class features. The proposed method is based on a neural network that uses the squared error as its error function, which is easy to calculate. The generated data is combined with the original minority class data, thereby reducing the proportion of generated data used and any effects due to differences between the generated and original data. We also expect that adding generated data will supplement the number of minority class features seen in the original data. Experiments evaluating our proposed method with artificial data show that it yielded a higher AUC value than that of undersampling. These results demonstrate that the proposed method can be an effective approach for addressing imbalanced datasets with small amounts of minority class data.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.
This website uses cookies
We use cookies to provide you with the best possible experience. They also allow us to analyze user behavior in order to constantly improve the website for you. Info about the privacy policy of IOS Press.