An Optimized Associative Classifier for Incremental Data Based On Non-Trivial Data Insertion

Abstract

Associative classification (AC) is an interesting approach in the domain of data mining which makes use of the association rules for building a classification system, which are easy for interpretation by the end user. The previous work [1] showed excellent performance in a static large data base but there existed a question of same performance when applied in an incremental data. Many of the Associative Classification methods have left the problem of data insertion and optimization unattended that results in serious performance degradation. To resolve this issue, we used new technique C-NTDI for building a classifier when there is an insertion of data that take place in a non-trivial fashion in the initial data that are used for updating the classification rules and thereafter to apply the PPCE technique for the generating of rules and further Proportion of Frequency occurrence count with BAT Algorithm (PFOCBA) is applied for optimizing the rules that are generated. The experiments were conducted on 6 different incremental data sets and we found that the proposed technique outperforms other methods such as ACIM, E-ACIM, Fast Update (FUP), Galois Lattice theory (GLT) and New Fast Update (NFUP) in terms of accuracy and time complexity.