Machine Learning based Missing Data Imputation Techniques

In the era of advanced machinery including sensor technology, massive amount of data than ever before is produced every day in a variety of different fields. In the process of handling huge data, some data values are missed due to various controllable or uncontrollable factors. The missing data problem is a real-life problem which found everywhere and that is why it has received huge attention in research in the past decades. It is a major hurdle in data analysis and research. Therefore, for better analysis of data there are various imputation methods to handle missing data varying from directly deleting the data which contains missing values or using traditional methods like mean, median and mode to fill the missing value to using complex deep learning and machine learning techniques to predict missing value. This paper first discusses different kinds of missing data and then analyzing various imputation techniques from traditional to modern approach for dealing with missing data. It also analyzes advantages and disadvantages of different imputation techniques to get a better perspective of these techniques.