Efficient data cleaning algorithm using decision tree classification model approach and modified new unique user identification algorithm using hashing techniques with a new error factor

Abstract

The study focuses on preprocessing techniques of web mining. Considering this scope, the study has proposed and implemented an effi-cient data cleaning and unique user identification algorithms. Previously proposed data cleaning algorithm is a generalized approach and lacked transparency. An appropriate model has to be used to implement the new data cleaning algorithm. Over analysis of various related studies and suggestions made by eminent experts, the study finalized decision tree classification model, and appropriate model to imple-ment the new data cleaning algorithm. Simplicity, ease in framing rules and ability to fragment complex decisions to solve a problem motivated to choose decision tree classification model to implement new data cleaning algorithm. Apart from this the study has also modified the previously proposed hash function, used to locate existing web users in web log server. A new error factor is introduced to remove memory address discrepancy. The modified hashing function along with binary search techniques is used to design the new unique user identification algorithm. Various experiments analysis is done using web log servers of eminent universities and colleges from United Arab Emirates and India. Results obtained prove the improved and better performances of the new rule based data cleaning and modified unique user identification algorithms.