RGNBC: Rough Gaussian Naïve Bayes Classifier for Data Stream Classification with Recurring Concept Drift

详细信息    查看全文

• 作者：D. Kishore Babu ; Y. Ramadevi ; K. V. Ramana
• 关键词：Data stream ; Recurring concept drift ; Naïve bayes ; Rough set theory ; Classification
• 刊名：Arabian Journal for Science and Engineering
• 出版年：2017
• 出版时间：February 2017
• 年：2017
• 卷：42
• 期：2
• 页码：705-714
• 全文大小：n>
• 刊物类别：Engineering
• 刊物主题：Engineering, general; Science, Humanities and Social Sciences, multidisciplinary;
• 出版者：Springer Berlin Heidelberg
• ISSN：2191-4281
• 卷排序：42

文摘

Due to the necessity of performing classification in streaming environments, researchers have developed various stream classification methods by handling concept drift. But, recurring concept drift is a challenging problem in data stream as the dimension of the data is not static over the period of time. By considering the recurring concept drift, this paper proposes a new classifier model, called Rough Gaussian Naïve Bayes Classifier (RGNBC) for the data stream classification. Here, two new contributions are made to handle the challenges of recurring concept drift. The first contribution is to utilize the rough set theory for detecting the concept drift. Then, gaussian naïve classifier is modified mathematically to handle the dynamic data without using the historic data. Also, the classification is performed using the posterior probability and the objective function which considers the multiple criteria. The proposed RGNBC model is experimented with two large datasets, and the results are validated against the existing MReC-DFS algorithm using sensitivity, specificity and accuracy. From the results, we proved that the proposed RGNBC model obtained the maximum accuracy of 74.5 % while compared with the existing algorithm.