Mining WebLog Data Using Enhanced Map-Reduce Based SOM Algorithm

Abstract

Web utilization mining is a procedure for finding  client route designs in web server to get web log data. These route designs are additionally scrutinized by different information mining strategies. The found route instances can be further used to recognize the successive transactions of the client, foreseeing the future demand of client, and so on which very much essential due to enormous developments in electronic trade sites.Through a colossal measure of web based shopping sites, it is important to see that what number of clients is really coming to the sites and making use of them. At the point when clients enter any online site, web logs are produced on the server. It is extremely significant to examine the web logs which help us in predicting the crisis inclines on electronic trade. These web based business sites create petabytes of log information consistently which isn't conceivable by conventional devices and methods to store and investigate valuable usage patterns. In this paper we propose a Hadoop based system  using Map Reduce methods which is extremelysolid for putting away such immense measure of information into HDFS and then unstructured logs are processed  through map reduced method to discover client conduct. Furthermore, in this paper we can likewise examine the log information using a combination of the supervised clustering algorithm  SOMwith Map Reduce paradigm. This proposed algorithm namedEnhanced Map-Reduced SOM is found to be 94% accurate.