Optimizing Online Time-Series Data Imputation Through Case-Based Reasoning

Abstract

When working with Intelligent Decision Support Systems (IDSS), data quality could compromise decisions and therefore, an undesirable behaviour of the supported system. In this paper, a novel methodology for time-series online data imputation is proposed. A Case-Based Reasoning (CBR) system is used to provide such imputation approach. The CBR principle (i.e., solving the current problem using past solutions to similar problems) may be applied to data imputation, using values from similar past situations to replace incorrect or missing values. To improve the performance of the data imputation process, optimal case feature weights are obtained using genetic algorithms (GA). The proposed methodology is validated with data obtained from a real Waste Water Treatment Plant (WWTP) process.