Data Cleansing for Data Quality Improvement in Data Mining

Good quality data are essential in data mining. Main tasks of data cleansing include missing value imputation and noisy value detection. We propose missing value imputation techniques that can automatically and accurately guesstimate missing values in data. We also propose a noise detection technique that automatically identifies data that are noisy or incorrect. If we clean up a dataset by our techniques then conventional data mining techniques such as a decision tree classifier are more likely to build better models.