Header menu link for other important links
CNB-MRF: Adapting Correlative Naive Bayes Classifier and MapReduce Framework for Big Data Classification
Published in Praise Worthy Prize
Volume: 11.0
Issue: 11.0
Pages: 1007.0 - 1015.0
Due to the continuous arrival of plenty of raw data, the big data classification is getting much attention in various fields, such as industry, medicine, and financial business. Due to the urgent need of big data classification in those fields, several standard classifiers are adapted to perform the big data classification, but they are not able to deal with these huge data problems. More commonly, Navie Bayes classifier is mostly preferred for big data classification because of its simple computation procedure. However, the main drawback behind this classifier is the independence hypothesis, in which the input data are conditionally independent of each other. To overcome this drawback, we propose a new classification algorithm using Correlative Naïve Bayes classifier and MapReduce framework (CNB-MRF). Here, the newly proposed correlation function with MapReduce framework is used to make the classification based on the dependent hypothesis to improve the classification performance. Here, the data classification can be easily performed for every new sample through the probability index table of the training data sample and the posterior probability of the testing data samples using MapReduce framework. Then, the classification by the proposed CNB-MRF classifier is performed using localization and skin dataset. It is concluded that, the proposed CNB-MRF classifier achieves a high classification accuracy of 74.77% and 61.35% for localization and skin dataset respectively, as compared with Naïve bayes classifier and MapReduce framework (NB-MRF).
About the journal
JournalInternational Review on Computers and Software
PublisherPraise Worthy Prize
Open AccessNo