Parallel association rule mining on semantic and big IoT data

Association Rule Mining (ARM) is an important machine learning technique because it can find associations or relationships between data items in large datasets. Different association rule algorithms have been studied by many researchers in traditional transactional data sets in which, all items have a unique relationship such as, 'buy'. In recent years, researchers have shown an increased interest in extracting association rules from semantic graphs (such as RDF datasets) instead of traditional data. On the other hand, in the field of the Internet of Things (IoT) and in many different domains, semantic data is daily increasing in large volumes. Therefore, we need scalable solutions to utilize IoT data for intelligent solutions. In this thesis, we studied the utilization of parallelized FP-growth algorithm on several different sematic IoT datasets from different domains, such as weather, traffic, and medicine. The results show that the scalable execution of semantic ARM algorithms produces the semantic association rules much faster. Keywords: Association rule mining, Scalable association rule mining, Semantic data, RDF, IoT, Machine learning, MapReduce.