Multivariate Imputation by N Neighbour Mean and Chained Equation for Time Series Missing Data

Missing or incomplete data introduces numerous challenges during data analysis. If there exists any incomplete data then algorithms used for analysis can produce incorrect results. Data with missing or incomplete values may have a negative effect on computation. Thereby leading to an incorrect result. Certain algorithms are incapable of handling missing data properly, while other methods produce an efficient output to calculate the missing values. Because different machine learning algorithms become less effective when given incomplete data, it is extremely important to manage missing data properly. It could be possible that the data set might contain some missing values for various features due to the following reasons data were maintained in the file, data got corrupted, etc. This paper proposed a method that makes use of N neighbors of the data that is missing and makes use of the machine learning model to tune the computed value so as to impute a correct value for the missing data when the missing data is spread across different features. The proposed model makes use of the spatial characteristics of the time series data in order to predict missing data with the least error. Through experimentation, the proposed model demonstrated a high degree of accuracy in predicting missing values, even when the gaps spanned multiple feature columns.