Improved Grey Relational Analysis for Model Validation

GRA (Grey Relational Analysis) is a typical time series similarity analysis method. However in model validation, it cannot satisfy the feature of monotonicity, and the result is lack of precision. Based on several similarity measurement criteria of time series data, traditional GRA method is developed and modified to satisfy normalization, symmetry and monotonicity. Case study shows that, improved GRA can produce a better similarity analysis result, which is in accordance with the result of TIC (Theil's Inequality Coefficient).