Associative Classification Algorithm And Its Systematic Implementation

With the development of the information of our society, the data exponentially increased. Quickly and efficiently to access to the most valuable information resource from the large amounts of data is of great significance in today's information age. Thus, there are very good prospects for development of the data mining techniques and tools. As an important means of intelligent decision-making system, the classification and prediction technology will play an important role in intelligent systems in the future. This paper did a comprehensive study in associative classification system technology, and a detailed theoretical explanation to the typical algorithm. Also, in order to solve the effect of the support and confidence thresholds, this paper has improved the existing associative classification algorithm, and proposed an association classification algorithm based on support and confidence thresholds optimization technology (Apriori _ TFP CMAR(HC)) , and also designed a rule extraction system based on the algorithm.First, the paper summarized the current development status of associative classification technology, as well as some correlative fields, such as association rules mining techniques.Then, the paper studied the association rule mining algorithms called Apriori_TFP and associative classification algorithm called CMAR, and through a combination of the key technologies of the two algorithms (improved data storage structure), the paper implemented a new classification algorithm called Apriori_TFP CMAR. This algorithm achieved better classification results in experiment, reducing the computation time and storage space occupied. Since then, current association classification algorithms always have a problem: because the support and confidence thresholds are always set by experiences, they usually affect the accuracy of classification based on association rules, and it is difficult to ensure that the classifier can always get the best accuracy. In order to solve the problem, this paper implemented the hill climbing search technology to obtain the support and confidence thresholds which can ensure the classifier to get the highest classification accuracy. Finally, the paper designed and implemented a rule extraction system based on Apriori_TFP CMAR (HC).The system also integrates with the data preprocessing functions, and gets a better experimental result.