Research on improvement of apriori algorithm based on marked transaction compression
Meng Xiao, Yong Yin, Yunyao Zhou, Shengzhi Pan · 2017
For a large sum of data collected and stored continually, it is more and more necessary to mine association rules from database, and the Apriori algorithm of association rules mining is the most classical algorithm of database mining and is widely used. However, Apriori algorithm has some disadvantages such as low efficiency of candidate item sets and scanning data frequently. Support and confidence are not enough to filter useless association rules. To recover the deficiencies, this paper puts forward an improved Apriori algorithm based on marked transaction compression, which optimizes the parameters of association rules (sup>1/2). Experiments show that this algorithm has much better capability than the original Apriori algorithm. After the second iteration of the algorithm, the candidate sets are reduced to 50%, the number of comparisons is reduced according to the tags, and the computational complexity of generating frequent item sets is decreased to 80%.