1Assistant Professor and PhD Scholar, Computer Engineering Department
2Assistant Professor, Information Technology Department, UV Patel College of Engineering, Mehsana, Gujarat, India
3Professor, Computer Engineering Department, Nirma Institute of Technology, Ahmedabad, Gujarat, India
*Email id: hit_r@rediffmail.com
Data pre-processing is an important task in knowledge discovery of database to analyse the quality of the data. In data mining, missing value imputation is a challenging issue confronted by data mining and machine learning. The way to recover the missing values which are all case deletion, single imputation, multiple imputation and iterative imputation. In this paper, we proposed new and effective single-imputation methods – Tuple Value-based Single Imputation (TVSI), Correlation-based Single Imputation (CORRSI) and Modified CORRSI (M-CORRSI) – to deal with missing values. Usually, data set has missing values in class attributes and conditional attributes. We proved effectiveness of our algorithms by J48 classification algorithm for class attributes in given data set. We compare our proposed single-imputation methods TVSI, CORRSI and M-CORRSI with Mean-based Single Imputation for efficiency and improvement.
Data mining, Pre-processing, Missing Values, Data Imputation, Classification
