Volume 18, No. 6, 2021
MISDSSP-Tree: A Novel Minimum Item Support Difference Based Pattern Tree Mining Approach For Mining Rare And Frequent Patterns
Keerti Shrivastava , Dr. Varsha Jotwani
Abstract
Life-threatening diseases are a major worry for many people in today's society. To limit the severity of the side effects of diseases, early identification and diagnosis are crucial. Rare Association Rule Mining (RARM), a method of computational intelligence, may be widely used in the study of illness. RARM study is dependent upon the assumption that all data to be mined is immediately accessible at beginning of the mining process. With the addition of new entries or deletion of old ones, medical databases in use might change over time. In addition, when the database is updated, the user may choose a new threshold for creating an appropriate collection of unusual association rules. A straightforward, but inept method may become the reconstruction of the whole mining algorithm from scratch, with each updated dataset and the revised threshold level for the present set of rare association rules. This paper presents an effective method in the context of rare patterns for 3 adverse diseases: hepatitis, breast cancer, & cardiovascular, for identification of symptoms & risky factors as described in the rare association rules. The Minimum Item Support Difference Single Scan Pattern-Tree (MISDSSP-tree) algorithm is used to calculate the support difference of each itemset. It checks the support count difference for each item with minimum support difference and compares it to the MIS value. MIS value satisfied itemset comprise frequent items & rare items. Experimental outcomes on real-life data sets demonstrated that presented MISDSSP-tree-algorithm increases performance against previous SSP-tree and ISSP-tree approaches by reducing the explosion of frequent itemset that contains frequent patterns and often comprise rare items. The MISDSSP-tree algorithm outperforms runtime and memory consumption. The relevance of the proposed approach over the usual strategy of repeatedly mining the complete updated database is shown via experimental analysis.
Pages: 3220-3240
Keywords: Data Mining, Association Rule Mining, Frequent and Rare Pattern Mining, Adverse Diseases, MIS, Support Difference, SSP-Tree.