Informatics Publishing Limited

Y. V. Siddartha Reddy ¹, K. P. Supreethi ¹

Affiliations
1 JNTU, Hyderabad, IN

Abstract

For classifications the fundamental technique is "Bayesian Classifier". However, it has been used programmatically. This paper throws light on programming Bayesian classifiers in query Language. Two classifiers are introduced in this paper. They are a classifier which uses K-means clustering based on class decomposition and Naïve Bayes Classifier. Scoring a data set and model computation are two tasks associated to those classifiers. The transformations of equations into query language queries are described including many query optimizations. The experiments are done to evaluate scalability, query optimizations and classification accuracy. The results revealed that the proposed Bayesian Classifier is more accurate when compared to Decision Trees and Naïve Bayes. Pivoting, renormalization, and horizontal layout of tables accelerated distance computation. Other observation is that query language implementation of Naïve Bayes is 4 times slower than its C++ counterpart. However, Bayesian classifier in SQL bestows high classification accuracy as it has linear scalability, and can effectively analyze large data sets.

Keywords

Query Optimization, k-Means, and Classification.

Full Text