Existential Probability Weighting Strategy to Reduce Search Space & Time for Big Data Mining

Abstract

Very huge quantity of data is continuously generated from a variety of different sources such as IT industries, internet applications, hospital history records, social media feeds etc. called as "Big Data". Mostly Data mining algorithms find the interesting patterns of data from the value-based database where the information is exact. It is not so easy to discover interesting patterns from big data. To abstain from squandering a ton of space & time in searching down frequent item uncertain big data, proposed approach permits clients to show their enthusiasm for terms of succinct anti-monotone constraint. MapReduce technique is used to mine frequent patterns. Two sets of map and reduce functions are used by proposed system to mine valid singleton and non-singleton patterns. In proposed work, UF-tree algorithm generates tree structure of dataset and UF-growth mines frequent itemsets recursively. To further reduce the search space and execution time in uncertain big data, proposed work gives importance to the frequency of items using weighting factors, and calculate expected support of item on the basis of weight. It reduces the nodes in the first level of tree, which leads to a reduction in the size of the tree and execution time.