An Enhanced Approach to Handle Missing Values in Heterogeneous Dataset

Generally, data mining (sometimes called data or knowledge discovery, knowledge extraction, knowledge discovery) is the process of analyzing huge voluminous data from different perspectives and summarizing it into the useful information. Hence data quality is much important to get the high quality pattern as result. Quality decisions ought to be based on quality data. Data quality is affected by the presence of missing values called holes because of various reasons. In order to make the database as complete by filling the holes with plausible value, variety of imputation methods have been developed. But they are limited to handle missing values in homogenous attributes only. Few of the existing systems uses the mixture kernel function for imputing missing values in mixed attribute datasets. In the proposed work, new imputation framework has been developed to handle missing values in heterogeneous datasets. Firstly pre-imputation is performed using ENI (Encapsidated Neighbour Imputation) approach followed by the application of Gaussian Kernel function to both continuous and discrete attributes. The proposed framework is tested with its competitors for various standard missing rates over bench dataset UCI repository. The behaviour of the framework proposed in this paper is studied using the parameter RMSE and concluded that it is behaving good. Keywords— Data Mining, Missing Value Imputation, Kernel