Data normalization for aggregating time series: the constrained min-max method

Abstract

This paper presents a method for normalizing data in time series, when variables have different measurement units and they must be aggregated (e.g., for constructing a composite index). The proposed method, denoted as “Constrained Min-Max Method”, normalizes the range of variables, similarly to the Min-Max method, but uses a common reference that allows to ‘centre’ them, as in the case of index numbers. A comparison with the traditional normalization methods is also shown.