Paper Type
Complete Research Paper
Description
Nowadays, companies rely more than ever on stored data to support decision making. However, outdated data may result in wrong decisions and economic losses. Thus, measuring data currency is extremely important. Existing metrics for currency either assume that their input parameters are given, or estimate them statistically, which may not always be possible in applications and especially in the context of big data. To address this issu, we propose a metric for currency based on expert estimations. The metric is modelled as a fuzzy inference system, which consists of a set of parallel IF-THEN rules with linguistic variables as inputs and output. It thus allows for a well-founded quantification of expert estimations and the consideration of both subjective and objective data. In addition to presenting our metric, we provide methods for estimating its input parameters (age of the considered attribute valu and its decline rate). Furthermore, we demonstrate how the fuzzy inference system and thus the metric can be initialised and applied. The presented approach serves as a first step in modelling expert estimations as input to data quality metrics in a well-defined and structured way.
A FUZZY METRIC FOR CURRENCY IN THE CONTEXT OF BIG DATA
Nowadays, companies rely more than ever on stored data to support decision making. However, outdated data may result in wrong decisions and economic losses. Thus, measuring data currency is extremely important. Existing metrics for currency either assume that their input parameters are given, or estimate them statistically, which may not always be possible in applications and especially in the context of big data. To address this issu, we propose a metric for currency based on expert estimations. The metric is modelled as a fuzzy inference system, which consists of a set of parallel IF-THEN rules with linguistic variables as inputs and output. It thus allows for a well-founded quantification of expert estimations and the consideration of both subjective and objective data. In addition to presenting our metric, we provide methods for estimating its input parameters (age of the considered attribute valu and its decline rate). Furthermore, we demonstrate how the fuzzy inference system and thus the metric can be initialised and applied. The presented approach serves as a first step in modelling expert estimations as input to data quality metrics in a well-defined and structured way.