ABSTRACT

It is very difficult to over-emphasize the benefits of accurate data. Errors in data are generally the most expensive aspect of data entry, costing the users even much more compared to the original data entry. Unfortunately, these costs are intangibles or difficult to measure. If errors are detected at an early stage then it requires little cost to remove the errors. Incorrect and misleading data lead to all sorts of unpleasant and unnecessary expenses. Unluckily, it would be very expensive to correct the errors after the data has been processed, particularly when the processed data has been converted into the knowledge for decision making. No doubt a stitch in time saves nine i.e. a timely effort will prevent more work at later stage. Moreover, time spent in processing errors can also have a significant cost. One of the major problems with automated data entry systems are errors. In this paper we discuss many well known techniques to minimize errors, different cleansing approaches and, suggest how we can improve accuracy rate. Framework available for data cleansing offer the fundamental services such as attribute selection, formation of tokens, selection of clustering algorithms, selection of eliminator functions etc.