Large-scale Data Mining: MapReduce and Beyond

Data are becoming available in unprecedented volumes. This difference in scale is difference in kind, presenting new opportunities. Map-reduce has drawn a lot of attention recent years for large-scale data processing and mining. In this tutorial, we introduce Map-reduce and its application and research in data mining. In particular, we want to answer the following questions: •What is Map-reduce and why do we need it for data mining? •What mining applications need Map-reduce? •What are the advantages and limitations using Map-Reduce? •How do you use Map-reduce? •What are other tools out there for large-scale data processing and mining? More specifically, this tutorial is organized into three parts: 1.MapReduce basic includes MapReduce programming model, system architecture, its OpenSource implementation Hadoop and its extensions such as HBase, Pig, Cascading, Hive. 2.MapReduce algorithms cover MapReduce implementation of standard data mining algorithms such as clustering (K-means), classification (k-NN, naive Bayes), graph mining (page rank). 3.MapReduce applications present the general applications of MapReduce that are beyond data mining, which include text processing, data warehousing.