Home // DATA ANALYTICS 2014, The Third International Conference on Data Analytics // View article

Prediction Model Framework for Imbalanced Datasets

Authors:
Maria Rossana de Leon
Eugene Rex Jalao

Keywords: Prediction model framework; Class imbalance problem; Data mining.

Abstract:
Generally, prediction requires significant and good quality input data that will give accurate prediction. However, real-data are often noisy, inconsistent, and imbalanced. If the classes are imbalanced, the class accuracy is unlikeable because the prediction tends to favor those in majority class since it has relatively significant class size. To resolve the imbalance problem, a resampling algorithm is proposed which improves the prediction accuracy of each class. The algorithm was tested in 4 different datasets each using different prediction and classification methodologies, such as Regression Analysis, Decision Tree, Rule Induction, and Artificial Neural Networks. Results show that the framework works in either methodologies and prediction accuracy generally improves after resampling. The framework was also compared to the existing sampling methodologies and results show that it is comparable with the ROS/RUS, but the resampling rate is minimized with the proposed framework.

Pages: 33 to 41

Copyright: Copyright (c) IARIA, 2014

Publication date: August 24, 2014

Published in: conference

ISSN: 2308-4464

ISBN: 978-1-61208-358-2

Location: Rome, Italy

Dates: from August 24, 2014 to August 28, 2013