Data Mining and Knowledge Engineering

Open Access Open Access  Restricted Access Subscription Access
Open Access Open Access Open Access  Restricted Access Restricted Access Subscription Access

Incremental Discretization for Naïve Bayes Learning with Optimum Binning


Affiliations
1 Charotar University of Science and Technology, Changa, Gujrat, India
2 Charotar University of Science and Technology Changa, Gujrat, India
3 Department of Computer Engineering, Dharamsinh Desai University, Nadiad, Gujarat, India
     

   Subscribe/Renew Journal


Incremental Flexible Frequency Discretization (IFFD) is a recently proposed discretization approach for Naïve Bayes (NB).IFFD performs satisfactory by setting the minimal interval frequency for discretized intervals as a fixed number. In this paper, we first argue that this setting cannot guarantee that the selecting MinBinSize is on always optimal for all the different datasets. So the performance of Naïve Bayes is not good in terms of classification error. We thus proposed a sequential search method for NB: named Optimum Binning. Experiments were conducted on 4 datasets from UCI machine learning repository and performance was compared between NB trained on the data discretized by OB, IFFD, and PKID.


Keywords

Discretization, Naïve Bayes, Optimum Binning.
User
Subscription Login to verify subscription
Notifications
Font Size

Abstract Views: 242

PDF Views: 3




  • Incremental Discretization for Naïve Bayes Learning with Optimum Binning

Abstract Views: 242  |  PDF Views: 3

Authors

Kamal Sutaria
Charotar University of Science and Technology, Changa, Gujrat, India
Amit Ganatra
Charotar University of Science and Technology Changa, Gujrat, India
Y. P. Kosta
Charotar University of Science and Technology, Changa, Gujrat, India
C. K. Bhensdadia
Department of Computer Engineering, Dharamsinh Desai University, Nadiad, Gujarat, India
Kruti Khalpada
Charotar University of Science and Technology, Changa, Gujrat, India

Abstract


Incremental Flexible Frequency Discretization (IFFD) is a recently proposed discretization approach for Naïve Bayes (NB).IFFD performs satisfactory by setting the minimal interval frequency for discretized intervals as a fixed number. In this paper, we first argue that this setting cannot guarantee that the selecting MinBinSize is on always optimal for all the different datasets. So the performance of Naïve Bayes is not good in terms of classification error. We thus proposed a sequential search method for NB: named Optimum Binning. Experiments were conducted on 4 datasets from UCI machine learning repository and performance was compared between NB trained on the data discretized by OB, IFFD, and PKID.


Keywords


Discretization, Naïve Bayes, Optimum Binning.