Research And Implement Of A Frequent Pattern List Based Associative Classification Algorithm

Mining association rules and classification technique are the two key techniquesin the fields of data mining researches. Associative classification algorithm made acombination of these two techniques which creates a new direction to build aclassification model. A large number of studies have shown that associativeclassification algorithm has better accuracy and adaptability than the traditionalclassification algorithm. But most of associative classification algorithms such as theclassical CMAR usually generate association rule set based on the FP-growth, theclassification model often stores the generated rule set with the tree structure whichcomplex and difficult to manage and store, so the efficiency and accuracy ofclassification is difficult to further improve.The main work of this paper is how to generate a more accurate association rulesin the stage of rule generation, and how to reduce the impact of the tree structure’scomplexity in classification model construction. In this regard, this paper proposes anew associative classification algorithm CBCFPL. New algorithm joins informationof classification in the FPL proposes CFP-list, then made a combination of frequentclosed itemsets mining and optimal rule set to construct classification based onCFP-list. Finally, the experiment proved CBCFPL with higher accuracy than thetraditional CBA and CMAR algorithms.Details in this paper are as follows:1) Describes the study background and significance of this paper, summarizesthe research situation and future trends of the associative classification in the field ofdata mining.2) Introduces the background knowledge of this paper.Topics include the introduction of association rule mining, classification miningand the related concepts of associative classification, describes the three classicalgorithm in details, and introduces the advantages and disadvantages in thealgorithm; 3) Analyzes the limitations of the FP-growth algorithm, introduce FPL thenimprove it.FP-growth algorithm storage and mining frequent pattern with FP-tree, thestructure of tree complex so that it is hard to management and storage. This paperintroduced FPL, then adds a suitable classification information in it, constructs andproposes a CFP-list. Compared with the FP-tree, given the features of the CFP-list;4) Proposes a new algorithm CBCFPL based on CFP-list.This algorithm constructs the classification model with CFP-list, then miningfrequent closed itemsets with the vector operation and trim the generated rule set onoptimal rule set. In this paper, we shows the flow chart and the pseudo-code of thekey steps.5) Conduct an experiment on CBCFPL and analyzed the result.We selected six data sets in UCI to conduct experiments to prove that thisalgorithm has higher classification accuracy than CBA and CMAR.The contents of this paper is an improvement to the traditional associativeclassification algorithm, mining association rules by table structure instead of the treestructure through the introduction of FPL, it reduces the complexity of the algorithmby constructing optimal rule set and the stategies of rule select, generates rules moreaccurate.This paper has a certain significance in research of efficiency improvementin associative classification of data mining.