Algqrithnis Of Mining Frequent Patterns For Uncertain Data Stream

Data streams model appears widely in many applications, the data is characterizedby fast, large-scale, real-time and single-pass accessing. Due to equipment accuracy,transmission loss, ambient interference, equipment failure, privacy protection andintegration between different systems and other reasons, uncertainty in the data streamis also widespread. Because of the existence and accuracy of the data in uncertain datastreams are expressed as the probability, the traditional data mining algorithms forcertain data can not meet the urgent requirement for uncertain data streams mining.Therefore, the combined data stream mining technology and uncertain data processingtechnology to design data mining algorithms for large-scale uncertain data streams hasbecome a research hot topic.After studying the problem of frequent pattern mining on uncertain data streams,we present a frequent pattern mining algorithm based on tilted time window foruncertain data stream. When dealing with frequent pattern mining, the data struct ofprefix tree UG-Tree can be used to compress and store of pattern information on theuncertain data streams. In order to reduce memory consumption and save executiontime in the mining process, pruning strategy is applied to delete the absolute sparsepattern in the prefix tree.The main work are as the follows:(1) For the characteristics of the uncertain data stream and application requirementsfor data mining, an uncertain data stream mining model is proposed,(2) We design a Synopsis Data Structure UG-Tree based on the prefix tree, and givean efficient pruning method for pruning absolute sparse pattern,(3) We propose an improved FP-Growth TOP-K frequent pattern miningalgorithm—UG-Miner for mining uncertain data streams TOP-K frequent patterns.The theoretical analysis and the simulation results show that algorithm caneffectively mine frequent patterns in uncertain data streams under the guarantee of thelimited memory consumption and real-time response.