Data Processing and deploying missing data algorithms to handle missing data in real time data of storage tank: A Cyber Physical Perspective

Abstract

Water forecasting is crucial for planning, designing the infrastructure, and also for operating and managing water supply systems. Forecasting in the computation unit plays a very significant role in Cyber-Physical System. Real-time monitoring of water flow rates information helps us to conserve water when it is needed the most. Hence, we summarize in this paper the first impediment in forecasting which is; handling missing data in the real-time monitoring system using the different imputation techniques such as k-Nearest Neighbor, Expectation-Maximization, Matrix Completion. The performance of the respective method is evaluated using traditional methods like Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). Missingness simply refers to the manner in which a demographic sample lacks data. The two types of missingness considered here are: 'missing at random' and 'missing at burst'. Using the different imputation methods, the validation is performed by computing for a range of 'missingness' varying from 0% to 60%. In our experiment setup, we synthetically omitted missing values continuously and randomly in the datasets for the rationale of imputing omitted values in the datasets. In this work, we are choosing the best-fitted model for our application. � 2019 IEEE.