AUTO-CDD: automatic cleaning dirty data using machine learning techniques
Jesmeen M. Z. H., Abid Hossen, J. Hossen, J. Emerson Raja, Bhuvaneswari Thangavel, S. Sayeed, Tawsif K.
Cleaning the dirty data has become very critical significance for many years, especially in medical sectors. This is the reason behind widening research in this sector. To initiate the research, a comparison between currently used functions of handling missing values and Auto-CDD is presented. The developed system will guarantee to overcome processing unwanted outcomes in data Analytical process; second, it will improve overall data processing. Our motivation is to create an intelligent tool that will automatically predict the missing data. Starting with feature selection using Random Forest Gini Index values. Then by using three Machine Learning Paradigm trained model was developed and evaluated by two datasets from UCI (i.e. Diabetics and Student Performance). Evaluated outcomes of accuracy proved Random Forest Classifier and Logistic Regression gives constant accuracy at around 90%. Finally, it concludes that this process will help to get clean data for further analytical process.
classification; data cleaning; dirty data; feature selection; gini index; random forest;