Mining of High-Utility Patterns in Big IoT Databases

Abstract

In general data mining, HUIM also known as high-utility itemset mining is an offshoot of frequent item set mining (FIM). HUIM is known to give more emphasis to many factors which can give HUIM a distinct edge over FIM. PHIUM, or Potential high-utility item set mining has been created to give intrinsic patterns in databases that tend to be uncertain. Despite most previous methods being highly effective and powerful miners, PHUIM needs to work fast. Most current mining techniques do not handle databases with extremely large number of records when performing HUIM. In this paper, we make the assumption that the dataset is bigger than a direct load into RAM could handle. Furthermore, the dataset is not of the size where modification or duplication is possible, and as such a MapReduce framework is created that can be used to handle datasets that fall into these categories. One of the main objectives of this research is to be able to reduce the frequency of database scans while simultaneously maximizing parallel processing. Using experimental analysis, our Hadoop based algorithm performs well to mine high utility itemsets from big databases.