A Data Quality Assessment Approach in the SmartWork Project's Time-series Data Imputation Paradigm.

The plethora of collected data streams of the SmartWork project's sensing system is often accompanied by missing values, yielding the need for estimating these missing values through imputation, which may prove unnecessary or computationally expensive in relation to the outcome. This work introduces a data quality assessment approach that allows for decision making regarding the need/efficiency of data completion in order to save system computational resources and ensure quality of imputed data. Preliminary validation of the proposed approach is performed by assessing the correlation between the proposed data quality assessment scores and the normalized mean square error of the imputation on various simulated missing patterns. The results reinforce our initial hypothesis that the suggested score is a suitable data quality indicator, correlating well with the potential errors introduced by imputation in the case of a given batch of input data.