Downloads: 141 | Views: 241
Research Paper | Computer Science & Engineering | India | Volume 5 Issue 7, July 2016 | Rating: 6.8 / 10
Improving Apriori Algorithm Using Shuffle Algorithm
Aasma Parveen | Shrikant Tiwari
Abstract: Data mining is the method of extracting interesting (non-trivial, embedded, previously indefinite and potentially useful) in sequence or patterns from large information repositories. Association mining aims to extract frequent patterns, interesting correlation, association or untailored structures among the sets of objects in the transaction files or from the other data repositories. It plays a vital role in spawning frequent item sets from large transaction databases. The discovery of remarkable organization relationship among business transaction records in many commercial decision making method such as catalog decision, cross-marketing, and loss-leader analysis. It is also used to excerpt hidden information from huge data sets. The Association Rule Mining algorithms such as Apriori, FP-Growth wants repetitive scans over the entire file. All the input/output overheads that are being generated during the frequent perusing process, entire file decreases the performance of CPU, memory and I/O overheads. In this paper we have proposed An Cohesive tactic of Parallel Processing and ARM for mining Association Rules on Generalized data set that is basically altered from all the preceding algorithms in that it use database in transposed form and database rearrangement is done using Parallel rearrangement algorithm (Shuffle Transpose) so to generate all important association rules number of passes essential is abridged and equaled various classical Association Rule Mining algorithms and topical procedures. The proposed Apriori algorithm has decreased the time complexity, by reducing the the processing time of the Transposition of the data sets. The comparison is done between the sequential and shuffle transposition using apriori algorithm which indicates the time difference of 28 seconds when the 100 X 100 matrix is considered, which was a very important aspect of the work. In the future, the work can be extended by parallelizing the algorithm for a communal nothing multiprocessor machine.
Keywords: Data Mining, Association Rule Mining ARM, Association rule, Apriori algorithm, Frequent patterns
Edition: Volume 5 Issue 7, July 2016,
Pages: 158 - 163