Association Rule Mining Based Video Classifier with Late Acceptance Hill Climbing Approach

Video classification is an essential step towards video perceptive. In recent years, the concept of utilizing association rules for classification emerged. This approach is more efficient and accurate than traditional techniques. Associative classifier integrates two data mining tasks such as association rule discovery and classification, to build a classifier for the purpose of prediction. The accuracy of classification will be influenced by the choice of appropriate values for whatever thresholds are used. In this paper, we present an effective video classification technique which employs the association rule mining and examine the effect of varying the support and confidence thresholds on the accuracy of the proposed algorithm. Instead of two stage associative classification method, Total from Partial Classification technique integrates the association rule discovery and classification in a single processing step to reduce the cost of pruning. TFPC uses two enumeration trees such as Partial support tree (P-Tree) and Total support tree (T-Tree). TFPC algorithm first loads the input data into the P-tree structure to reduce the storage requirements. In the second stage, T-tree is used to hold the information regarding to frequent items. We proposed Late Acceptance Hill-Climbing (LAHC) method is to find the effective support and confidence coupled with TFPC. The LAHC method accepts the candidates with cost function better than the cost solution which were the current several iterations before. Experimental results show that the performance of classification accuracy can be significantly improved.